{
 "schema": "boar-publication/facts",
 "schema_version": "1.0.0",
 "generated": "2026-10-25",
 "license": "CC BY 4.0 (text and data); see /CITATION.cff",
 "count": 2138,
 "chapters": [
  {
   "id": "question",
   "n": 1,
   "title": "The question",
   "span": "2026-09-17 → 09-30",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "ruler",
   "n": 2,
   "title": "Building a ruler",
   "span": "2026-10-01 → 10-03",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "floor-trap",
   "n": 3,
   "title": "The floor trap",
   "span": "2026-10-02 → 10-07",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "builds",
   "n": 4,
   "title": "Seven frozen builds",
   "span": "2026-10-03 → 10-07",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "walls",
   "n": 5,
   "title": "The walls",
   "span": "2026-10-03 → 10-06",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "deep-tier",
   "n": 6,
   "title": "The deep tier",
   "span": "2026-10-02, 10-06 → 10-10; dropped from delivery 10-21",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "phone",
   "n": 7,
   "title": "A phone on a desk, and in a hand",
   "span": "2026-10-02 → 10-22",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "agent-lab",
   "n": 8,
   "title": "The agent lab",
   "span": "2026-10-01 → 10-24",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "packs",
   "n": 9,
   "title": "Knowledge packs",
   "span": "2026-10-12 → 10-14",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "measurer",
   "n": 10,
   "title": "Measuring the measurer",
   "span": "2026-10-02 → 10-21",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "small-engine",
   "n": 11,
   "title": "A smaller engine for 8 GB phones",
   "span": "2026-10-18 → 10-21",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "new-ruler",
   "n": 12,
   "title": "A second ruler (bench v0.4)",
   "span": "2026-10-19 → 10-24",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "delivery",
   "n": 13,
   "title": "Delivery, not coverage",
   "span": "2026-10-19 → 10-23",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "closure",
   "n": 14,
   "title": "Closing the research: what a small model needs",
   "span": "2026-10-23 → 10-24",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "submission",
   "n": 15,
   "title": "One simple app: LabRat",
   "span": "2026-10-15 → 10-25",
   "note": "editorial order fixed 2026-10-24; title renamed 2026-10-25 (BoarLite → LabRat; content/narrative.md, 18 chapters)"
  },
  {
   "id": "honest-number",
   "n": 16,
   "title": "Where it stands (live)",
   "span": "2026-10-05 → 10-25",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "handoff",
   "n": 17,
   "title": "Open fronts and hand-off",
   "span": "2026-10-14 →",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  },
  {
   "id": "reproduce",
   "n": 18,
   "title": "Reproduce it (appendix)",
   "span": "—",
   "note": "editorial order and title fixed 2026-10-24 (content/narrative.md, 18 chapters)"
  }
 ],
 "layers": [
  {
   "id": "curious",
   "title": "Curious",
   "promise": "The story and the headline numbers, no jargon."
  },
  {
   "id": "tech",
   "title": "Tech people",
   "promise": "How the system and the method work, with the numbers that drive each decision."
  },
  {
   "id": "ai",
   "title": "AI professionals",
   "promise": "Every number with set, status, judge, n, seeds and CI; power, noise and threats to validity."
  }
 ],
 "notes": [
  "Always read raw together with floor-adjusted, and note set, set_status, judge, n and seeds. The 0.50 bar was not reached under the Claude judge on held-out data.",
  "superseded_by (optional, list of F-ids) and superseded_note mark an older fact whose standing a later row changed; the old row is kept as it was."
 ],
 "facts": [
  {
   "id": "F-001",
   "kind": "fact",
   "claim": "The frozen 4B claim build on the bounty-style set (V-para) delivers about a quarter of the frontier reference's value above \"nothing\", and loses every head-to-head.",
   "value": "raw 0.413 (0.384–0.445); floor-adjusted 0.261 (0.224–0.300); win/tie/loss 0/0/51 in both seeds",
   "ci": "95 % item bootstrap",
   "n": "51",
   "seeds": "2 (42, 43)",
   "set": "V-para",
   "set_status": "held-out for this run (cut 3 on 2026-10-05, cut 6 on 2026-10-06); dev-style for everything measured after 2026-10-06",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-06",
   "source_public": "boar-research:reports/BAR-TABLE.md:35; boar-research:reports/SET-STATUS.md:15; boar-research:README.md:25; boar-research:reports/BAR-TABLE.md",
   "caveat": "Cut 7 = cut 6 on this set (the FACTS table fires 0 times on it). Public BAR-TABLE prints CI high 0.300; one claim draft prints 0.299 (rounding).",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "canon:k-vpara",
    "F-results-01",
    "F-results-02",
    "F-system-38",
    "F-chron-early-63",
    "F-process-07"
   ]
  },
  {
   "id": "F-002",
   "kind": "fact",
   "claim": "The same 4B claim build on the second set (v0.2, later burnt to dev).",
   "value": "raw 0.512 (0.478–0.549); floor-adjusted 0.388 (0.346–0.434); win/tie/loss 0/0/110 and 0/1/109 (per seed)",
   "ci": "95 % item bootstrap",
   "n": "110",
   "seeds": "2",
   "set": "v0.2",
   "set_status": "held-out at run time; burnt → dev by ruling 2026-10-06",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-06",
   "source_public": "boar-research:reports/BAR-TABLE.md:53; boar-research:README.md:26",
   "caveat": "Claim draft v4.2.4 (§0 and Appendix A) carries the cut-3 values 0.513 / 0.389; the cut-6 = cut-7 values here are the ones to publish.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "canon:k-v02",
    "F-results-04",
    "F-results-05",
    "F-system-39"
   ]
  },
  {
   "id": "F-003",
   "kind": "fact",
   "claim": "A system that declines every question already scores about 0.31 on the raw ratio, because the rubric never goes below 1 of 5.",
   "value": "always-decline raw ratio D = 0.306 (v0.2-weighted 0.307)",
   "ci": null,
   "n": "79 judged declines over 7 runs",
   "seeds": null,
   "set": "v0.1 night runs",
   "set_status": "dev",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-02",
   "source_public": "boar-bench:bench/v0.3/plan/analysis-plan.md:24; boar-research:reports/BAR-TABLE.md:14; boar-research:README.md",
   "caveat": "This is why the floor-adjusted ratio, not the raw ratio, is the headline.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "canon:decline-floor",
    "F-process-06"
   ]
  },
  {
   "id": "F-004",
   "kind": "fact",
   "claim": "Bar status: no configuration reaches 0.50 on the primary (floor-adjusted) reading; the formal raw rule is met only on development data by the deep tier with thinking.",
   "value": "best floor-adjusted on V-para = 0.436 (CI high 0.472), short by 0.064; formal raw rule (CI low > 0.50) met by deep + thinking on dev data only (desktop; the v0.2 leg of the 512 budget is 1 seed); 0 wins on every Claude-judged row of V-para and v0.2",
   "ci": "see F-28",
   "n": "51 / 110",
   "seeds": "see rows",
   "set": "V-para; v0.2",
   "set_status": "dev / dev-style",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:80; boar-research:README.md:42; boar-research:reports/BAR-TABLE.md:73,80; boar-research:reports/BAR-TABLE.md:66-80",
   "caveat": "Never present as \"bar reached\". Under the Claude judge on held-out data the bar was not reached.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "canon:bar-status",
    "F-method-29",
    "F-results-37",
    "F-chron-early-107"
   ]
  },
  {
   "id": "F-005",
   "kind": "fact",
   "claim": "Swapping in the 35B mixture-of-experts deep tier (no thinking) raises the bounty-style score.",
   "value": "raw 0.467 (0.439–0.494); floor-adjusted 0.329 (0.294–0.362); win 0/51; desktop p50 ≈ 33 s",
   "ci": "95 % item bootstrap",
   "n": "51",
   "seeds": "2",
   "set": "V-para",
   "set_status": "dev-style (measured after 2026-10-06)",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-08",
   "source_public": "boar-research:reports/BAR-TABLE.md:37; boar-research:README.md:27",
   "caveat": "Desktop only; the deep tier has never run on a phone.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "canon:deep-vpara",
    "F-results-11",
    "F-system-40",
    "F-chron-early-90"
   ]
  },
  {
   "id": "F-006",
   "kind": "fact",
   "claim": "Deep tier (no thinking) on v0.2.",
   "value": "raw 0.568 (0.536–0.602); floor-adjusted 0.457 (0.417–0.500); win 0/110",
   "ci": "95 % item bootstrap",
   "n": "110",
   "seeds": "2",
   "set": "v0.2",
   "set_status": "dev (burnt)",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-08",
   "source_public": "boar-research:reports/BAR-TABLE.md:55",
   "caveat": null,
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "canon:deep-v02",
    "F-results-13",
    "F-chron-early-91"
   ]
  },
  {
   "id": "F-007",
   "kind": "fact",
   "claim": "Letting the deep tier think for up to 1024 tokens before answering.",
   "value": "raw 0.542 (0.513–0.571); floor-adjusted 0.422 (0.385–0.459); win 0/51; desktop p50 ≈ 143 s",
   "ci": "95 % item bootstrap",
   "n": "51",
   "seeds": "2",
   "set": "V-para",
   "set_status": "dev-style",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-08",
   "source_public": "boar-research:reports/BAR-TABLE.md:38",
   "caveat": null,
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "canon:think1024-vpara",
    "F-results-14"
   ]
  },
  {
   "id": "F-008",
   "kind": "fact",
   "claim": "Thinking is the largest single measured gain on the bounty-style set.",
   "value": "thinking (1024) vs no thinking: raw +0.075 (+0.040…+0.110); floor-adjusted +0.093 (+0.050…+0.137)",
   "ci": "paired 95 % bootstrap",
   "n": "51",
   "seeds": "2",
   "set": "V-para",
   "set_status": "dev-style",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-08",
   "source_public": "boar-research:reports/BAR-TABLE.md:106",
   "caveat": null,
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "canon:think-gain",
    "F-results-15"
   ]
  },
  {
   "id": "F-009",
   "kind": "fact",
   "claim": "Recommended deep configuration (35B mixture-of-experts + 512-token thinking budget) on the bounty-style set.",
   "value": "raw 0.553 (0.524–0.581; with seed noise 0.513–0.592); floor-adjusted 0.436 (0.401–0.472; with seed noise 0.387–0.486); win 0/51; desktop p50 ≈ 79 s",
   "ci": "95 % item bootstrap; common-mode CI adds seed noise",
   "n": "51",
   "seeds": "2",
   "set": "V-para",
   "set_status": "dev-style",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-09",
   "source_public": "boar-research:reports/BAR-TABLE.md:39; boar-research:reports/BAR-TABLE.md:126; boar-research:README.md:28",
   "caveat": "Chosen as the best of ≥ 8 deep configurations on this same set: expect a winner's-curse discount of about 0.01–0.02. QUEUE prints raw 0.552; the two-seed aggregate is 0.553. Desktop only.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "canon:h512-vpara",
    "F-results-18",
    "F-results-19",
    "F-system-41",
    "F-chron-late-03"
   ]
  },
  {
   "id": "F-010",
   "kind": "fact",
   "claim": "A 512-token thinking budget matches 1024 in quality at about half the time.",
   "value": "512 vs 1024 on V-para: raw +0.011 (−0.010…+0.031), floor-adjusted +0.014 (−0.012…+0.039), ×0.55 time; on v0.2 −0.005 (−0.020…+0.009), ×0.62 time",
   "ci": "paired 95 % bootstrap",
   "n": "51 / 110",
   "seeds": "2 (V-para) / 1 (v0.2)",
   "set": "V-para; v0.2",
   "set_status": "dev-style / dev",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-09",
   "source_public": "boar-research:reports/CLOSED.md:39; boar-research:reports/BAR-TABLE.md:107; boar-research:reports/CLOSED.md#R20",
   "caveat": "Verdict label: CHEAPER-PARITY (an equivalence, not a gain).",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "canon:budget512",
    "F-results-25",
    "F-results-26",
    "F-system-44"
   ]
  },
  {
   "id": "F-011",
   "kind": "fact",
   "claim": "A 256-token budget is cheaper but measurably worse.",
   "value": "raw 0.514 (0.483–0.544); floor-adjusted 0.387 (0.349–0.426); vs 1024 raw −0.028 (−0.056…−0.003); p50 ≈ 55 s",
   "ci": "paired 95 % bootstrap",
   "n": "51",
   "seeds": "2",
   "set": "V-para",
   "set_status": "dev-style",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-09",
   "source_public": "boar-research:reports/BAR-TABLE.md:41; boar-research:reports/CLOSED.md:40",
   "caveat": "Keeps ≈ 63 % of the thinking gain at 40 % of the time.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "canon:budget256",
    "F-results-27",
    "F-chron-early-97"
   ]
  },
  {
   "id": "F-012",
   "kind": "fact",
   "claim": "Three seeds of the recommended deep configuration show how much one run can move.",
   "value": "raw mean 0.550, SD 0.020 (0.533 / 0.572 / 0.544); floor-adjusted mean 0.432, SD 0.025 (0.412 / 0.460 / 0.425); 0 wins each seed",
   "ci": "pooled CI UNKNOWN (aggregate file empty)",
   "n": "51",
   "seeds": "3 (42–44)",
   "set": "V-para",
   "set_status": "dev-style",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:40; boar-research:reports/RULE-CHANGES.md:16",
   "caveat": null,
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "canon:h512-3seed",
    "F-method-33",
    "F-results-20",
    "F-chron-early-103",
    "F-chron-late-05"
   ]
  },
  {
   "id": "F-013",
   "kind": "fact",
   "claim": "Recommended deep configuration on v0.2: the first point above 0.50 floor-adjusted on any set, with a CI that straddles 0.50.",
   "value": "raw 0.608 (0.575–0.641); floor-adjusted 0.508 (0.467–0.550); win/tie/loss 0/1/109; desktop p50 ≈ 82 s",
   "ci": "95 % item bootstrap",
   "n": "110",
   "seeds": "1 (42)",
   "set": "v0.2",
   "set_status": "dev (burnt)",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-09",
   "source_public": "boar-research:reports/BAR-TABLE.md:57; boar-research:README.md:29",
   "caveat": "One seed, development data, desktop only.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "canon:h512-v02",
    "F-results-23",
    "F-system-42",
    "F-chron-early-98",
    "F-chron-late-04"
   ]
  },
  {
   "id": "F-014",
   "kind": "fact",
   "claim": "Deep tier with 1024-token thinking on v0.2.",
   "value": "raw 0.612 (0.581–0.644); floor-adjusted 0.514 (0.475–0.554); 0 wins",
   "ci": "95 % item bootstrap",
   "n": "110",
   "seeds": "2",
   "set": "v0.2",
   "set_status": "dev (burnt)",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-09",
   "source_public": "boar-research:reports/BAR-TABLE.md:56",
   "caveat": null,
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "canon:t1024-v02",
    "F-results-16"
   ]
  },
  {
   "id": "F-015",
   "kind": "fact",
   "claim": "Anchor: a small frontier model with no web access (Claude Haiku 5.5) on the bounty-style set.",
   "value": "raw 0.588 (0.559–0.615); floor-adjusted 0.48 (CI unknown)",
   "ci": "raw 95 % item bootstrap; floor-adjusted CI never computed",
   "n": "51",
   "seeds": "1",
   "set": "V-para",
   "set_status": "held-out at the time (2026-10-05)",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-05",
   "source_public": "boar-research:reports/BAR-TABLE.md:44",
   "caveat": "The best deep configuration (0.436, CI high 0.472) does not reach this anchor either.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "canon:haiku-anchor",
    "F-results-09"
   ]
  },
  {
   "id": "F-016",
   "kind": "fact",
   "claim": "Win-rate against the frontier reference: zero on both bounty-relevant sets under the Claude judge; a few isolated wins exist only on the public dev set.",
   "value": "0 wins on every row and seed of V-para and v0.2 (Opus); v0.1 dev: 1/57 under Opus for the deep R450, LES and 4B rows; 2/57 under Haiku on the 12 GB-class phone run",
   "ci": null,
   "n": "51 / 110 / 57",
   "seeds": "all",
   "set": "V-para; v0.2; v0.1",
   "set_status": "dev-style / dev / public dev",
   "judge": "Claude Opus; Claude Haiku L1 (phone run)",
   "date": "2026-10-14",
   "source_public": "boar-research:reports/BAR-TABLE.md:80; boar-research:arms/device-x6/DEVICE-X6.md:27",
   "caveat": "Public text that says \"0 everywhere\" (boar-research README.md:42, boar-bench TRACE-FORMAT.md:70) overstates the scope.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "canon:winrate-scope"
   ]
  },
  {
   "id": "F-017",
   "kind": "fact",
   "claim": "A judge from a different model family scores everything higher on V-para, but keeps the ranking and the near-zero win-rate.",
   "value": "4B: raw 0.497 (0.449–0.547), floor-adjusted 0.361; deep + think 512: raw 0.667 (0.621–0.711), floor-adjusted 0.577 (0.519–0.632); deep − 4B +0.171 (+0.123…+0.219) vs +0.139 under Claude; per-item Spearman 0.88–0.92; win-rate 0.00–0.01",
   "ci": "95 % item bootstrap (4B floor-adjusted CI not in the sources read)",
   "n": "51 (408 judge calls)",
   "seeds": "2",
   "set": "V-para",
   "set_status": "dev-style",
   "judge": "cross-family: OpenAI gpt-5.6-sol (via the Codex CLI) (post-hoc, not pre-registered)",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "Post-hoc and not pre-registered; it is the only point above 0.50 floor-adjusted on V-para and exists only under the cross-family judge; the lower-judge rule keeps the Claude number as headline.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "canon:xfam-vpara",
    "F-results-46",
    "F-results-47",
    "F-results-48",
    "F-chron-late-76",
    "F-chron-late-79"
   ]
  },
  {
   "id": "F-018",
   "kind": "fact",
   "claim": "The cross-family judge on v0.2.",
   "value": "4B: raw 0.617 (0.567–0.666), floor-adjusted 0.517 (2 seeds); deep + think 512: raw 0.714 (0.666–0.759), floor-adjusted 0.639 (0.579–0.695) (1 seed); deep − 4B +0.097 vs +0.096 under Claude; Spearman 0.90–0.97; win-rate 2.7 % / 6.4 % vs 0 % under Claude",
   "ci": "95 % item bootstrap",
   "n": "110 (660 judge calls)",
   "seeds": "4B 2, deep 1",
   "set": "v0.2",
   "set_status": "dev (burnt)",
   "judge": "cross-family: OpenAI gpt-5.6-sol (via the Codex CLI) (post-hoc, not pre-registered)",
   "date": "2026-10-17",
   "source_public": null,
   "caveat": "Post-hoc. The Claude judge is the stricter family by ≈ 0.10 on both sets; the offset is stable across sets and arms.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "canon:xfam-v02",
    "F-results-49",
    "F-results-50",
    "F-chron-late-77"
   ]
  },
  {
   "id": "F-019",
   "kind": "fact",
   "claim": "Re-judging the exact same answers barely moves the score: judge noise is negligible next to seed noise.",
   "value": "σ_judge ≈ 0.003 raw / 0.004 floor-adjusted; per-item r 0.96; same winner 51/51",
   "ci": "n 51; 102 calls",
   "n": "51 (102 Opus calls)",
   "seeds": "1 (re-judge of s42)",
   "set": "V-para",
   "set_status": "dev-style",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-11",
   "source_public": "boar-research:reports/BAR-TABLE.md:123; boar-research:reports/RULE-CHANGES.md:23",
   "caveat": null,
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "canon:judge-retest",
    "F-results-45",
    "F-process-30"
   ]
  },
  {
   "id": "F-020",
   "kind": "fact",
   "claim": "Seed noise: the same deep configuration run with two random seeds differs by more than many tested \"levers\".",
   "value": "s42 vs s43, same configuration: Δ −0.039 (−0.072…−0.011) raw; house rule since 2026-10-16: a one-seed deep-tier Δ below 0.05 on V-para is seed noise (raised from 0.03)",
   "ci": "paired 95 % bootstrap",
   "n": "51",
   "seeds": "2",
   "set": "V-para",
   "set_status": "dev-style",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "The public boar-research README.md:77 still says 0.03.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "canon:seed-noise",
    "F-method-34",
    "F-results-21",
    "F-chron-late-87"
   ]
  },
  {
   "id": "F-021",
   "kind": "fact",
   "claim": "Part of the gap to the reference is length: shortening the references to 80 words raises the deep tier's score, mostly through completeness and usefulness.",
   "value": "deep + think 512 floor-adjusted 0.412 → 0.514 (0.456–0.578) with ≤ 80-word references; paired +0.102 (+0.068…+0.144); ≈ 17 % of the gap is length, 83 % substance",
   "ci": "paired 95 % bootstrap",
   "n": "51",
   "seeds": "1 (42)",
   "set": "V-para",
   "set_status": "dev-style",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-11",
   "source_public": "boar-research:reports/BAR-TABLE.md:98; boar-research:reports/BAR-TABLE.md:97-98; boar-research:reports/BAR-TABLE.md:97",
   "caveat": "Diagnostic with altered references; not a bar reading.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "canon:lengthshare",
    "F-method-32",
    "F-chron-late-07"
   ]
  },
  {
   "id": "F-022",
   "kind": "fact",
   "claim": "On correctness alone the deep tier is much closer to the reference than on the full rubric.",
   "value": "floor-adjusted correctness-only: V-para 0.660 (0.608–0.712); v0.2 0.743 (0.681–0.805)",
   "ci": "95 % item bootstrap",
   "n": "51 / 110",
   "seeds": "2 / 1",
   "set": "V-para; v0.2",
   "set_status": "dev-style / dev",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:89; boar-research:reports/BAR-TABLE.md:91",
   "caveat": "Single-axis diagnostic, not a bar reading.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "canon:correctness-only"
   ]
  },
  {
   "id": "F-023",
   "kind": "fact",
   "claim": "On a real 12 GB-class phone, in airplane mode, the frozen build gives the same judged quality as on the desktop.",
   "value": "phone raw 0.529 (0.481–0.575) vs desktop 0.529; floor-adjusted 0.403 vs 0.402; paired −0.000 (−0.026…+0.028); 72/72 answered; fast-tier p50 30.7 s; app memory (PSS) peak 5.88 GB during model load, then ≤ 4.23 GB",
   "ci": "paired 95 % bootstrap",
   "n": "57",
   "seeds": "1 run",
   "set": "v0.1",
   "set_status": "public dev",
   "judge": "Claude Haiku 5.5-low (L1 triage judge, N = 4, both orders)",
   "date": "2026-10-14",
   "source_public": "boar-research:arms/device-x6/DEVICE-X6.md:15-18; boar-research:arms/device-x6/DEVICE-X6.md:27; boar-research:arms/device-x6/DEVICE-X6.md:27-30; boar-research:arms/device-x6/DEVICE-X6.md:27-28; boar-research:arms/device-x6/DEVICE-X6.md:13",
   "caveat": "Haiku-judged level; not comparable with Opus-judged ratios. The device has 10.7 GiB usable RAM.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "canon:x6-parity",
    "F-results-69",
    "F-chron-late-52",
    "F-chron-late-53"
   ]
  },
  {
   "id": "F-024",
   "kind": "fact",
   "claim": "Earlier, on an 8 GB phone, cut 6 also matched the desktop.",
   "value": "phone 0.561 (0.507–0.617) vs desktop 0.558 / 0.560; 73/73 answered; PSS max 4.41 GiB",
   "ci": "95 % item bootstrap",
   "n": "57",
   "seeds": "1 run",
   "set": "v0.1",
   "set_status": "public dev",
   "judge": "Claude Opus (judge.v1-med)",
   "date": "2026-10-06",
   "source_public": "boar-research:README.md:54; boar-research:README.md",
   "caveat": "Opus-judged; not comparable in level with the Haiku-judged 12 GB-class run.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "canon:a15-parity",
    "F-results-73",
    "F-system-30",
    "F-chron-early-70"
   ]
  },
  {
   "id": "F-025",
   "kind": "fact",
   "claim": "One matching knowledge pack helps the small model a little on questions written from that pack.",
   "value": "Opus +0.042 (+0.024…+0.062); Haiku +0.032 (+0.016…+0.050)",
   "ci": "paired 95 % bootstrap",
   "n": "180",
   "seeds": "1",
   "set": "180 pack self-tests",
   "set_status": "dev (agent-written from pack text)",
   "judge": "Claude Opus (L2) and Claude Haiku (L1)",
   "date": "2026-10-13",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:8",
   "caveat": "Pack-matched dev questions, not a bounty-bar measurement.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "canon:pack-single",
    "F-results-77"
   ]
  },
  {
   "id": "F-026",
   "kind": "fact",
   "claim": "Installing all packs at once crowds out the right passage and costs a little quality.",
   "value": "Opus −0.019 (−0.035…−0.003); Haiku −0.016 (−0.033…−0.001)",
   "ci": "paired 95 % bootstrap",
   "n": "180",
   "seeds": "1",
   "set": "180 pack self-tests",
   "set_status": "dev",
   "judge": "Claude Opus (L2) and Claude Haiku (L1)",
   "date": "2026-10-13",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:11-12; boar-research:reports/REPORT-PACKS-2026-10-13.md:107,111",
   "caveat": null,
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "canon:pack-crowding",
    "F-results-81",
    "F-system-70",
    "F-chron-late-30"
   ]
  },
  {
   "id": "F-027",
   "kind": "fact",
   "claim": "Ceiling test: handing the model the exact fact-bearing chunk shows the headroom is retrieval, not reading.",
   "value": "inject fact chunk (ADD) +0.095 (+0.070…+0.122); fact chunk only (ONLY) +0.115 (+0.088…+0.142)",
   "ci": "paired 95 % bootstrap",
   "n": "180",
   "seeds": "1",
   "set": "180 pack self-tests",
   "set_status": "dev",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-13",
   "source_public": "boar-research:arms/oracle-st/ORACLE-ST.md:3; boar-research:reports/REPORT-PACKS-2026-10-13.md:19; boar-research:arms/oracle-st/ORACLE-ST.md:3,8",
   "caveat": "Oracle (cheating) cell: an upper bound, not a deployable result.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "canon:oracle-chunk",
    "F-results-83",
    "F-system-72",
    "F-chron-late-35"
   ]
  },
  {
   "id": "F-028",
   "kind": "fact",
   "claim": "The one retrieval change that converted: append the retrieved article's neighbouring chunks plus a few vector-search candidates, under a 600-token budget (HEADPASS-X2b).",
   "value": "Opus +0.066 (+0.043…+0.091); Haiku +0.056; fact chunk present in context 35 → 102 of 180",
   "ci": "paired 95 % bootstrap",
   "n": "180",
   "seeds": "1",
   "set": "180 pack self-tests",
   "set_status": "dev (designed on the same set)",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-14",
   "source_public": "boar-research:arms/headpass-st/HEADPASS-ST.md:57; boar-research:arms/headpass-st/HEADPASS-ST.md:89; boar-research:reports/REPORT-PACKS-2026-10-13.md:21; boar-research:arms/headpass-st/HEADPASS-ST.md:57,89; boar-research:arms/headpass-st",
   "caveat": "LEVER on pack-matched dev data only; inert on the claim kit (ceiling ≈ 0.011) and not a bounty-bar lever.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "canon:x2b",
    "F-results-86",
    "F-process-20"
   ]
  },
  {
   "id": "F-029",
   "kind": "fact",
   "claim": "The 35B deep tier gains nothing measurable from packs; its advantage over the 4B with the same pack is model knowledge.",
   "value": "deep with vs without pack Δ −0.002 (−0.020…+0.016); deep − 4B with the same pack +0.054 (+0.034…+0.073)",
   "ci": "paired 95 % bootstrap",
   "n": "180",
   "seeds": "1",
   "set": "180 pack self-tests",
   "set_status": "dev",
   "judge": "Claude Haiku 5.5-low (L1 triage judge, N = 4, both orders)",
   "date": "2026-10-13",
   "source_public": "boar-research:arms/packs/SELFTEST-DEEP.md:4; boar-research:arms/packs/SELFTEST-DEEP.md:18; boar-research:reports/REPORT-PACKS-2026-10-13.md:10; boar-research:reports/REPORT-PACKS-2026-10-13.md:209",
   "caveat": null,
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "canon:deep-packs",
    "F-results-79",
    "F-results-80",
    "F-chron-late-29"
   ]
  },
  {
   "id": "F-030",
   "kind": "fact",
   "claim": "Long conversations do not degrade answers up to turn 15.",
   "value": "turns 6–15, in-session vs standalone: 1.136 (0.993–1.320); self-contained items 1.017 (0.950–1.095); latency ≈ 22 s vs 8 s at turns 14–15",
   "ci": "95 % bootstrap",
   "n": "80 agent-written questions",
   "seeds": "1",
   "set": "LONGSESSION dev sessions",
   "set_status": "dev (agent-written)",
   "judge": "Claude Haiku 5.5-low (L1 triage judge, N = 4, both orders)",
   "date": "2026-10-14",
   "source_public": "boar-research:arms/longsession/LONGSESSION.md:68; boar-research:arms/longsession/LONGSESSION.md:81; boar-research:arms/longsession/LONGSESSION.md:68-80",
   "caveat": "The reference here is BOAR's own standalone answer, not the frontier reference: a ratio of ratios, not a bar number.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "canon:longsession",
    "F-results-75",
    "F-chron-late-56"
   ]
  },
  {
   "id": "F-031",
   "kind": "fact",
   "claim": "Four days of an unattended agent research loop produced one dev-only lever.",
   "value": "133 ticks (28 idle), 2026-10-11 12:10 → 10-14 19:09; ≈ 10.5k judge calls (Opus 4,851 + Haiku 5,660) plus 1,491 Haiku writer/generation calls; Mac mini 44.4 h + MacBook 18.1 h of generation; ≈ 31 h avoidable",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "canon:loop-economics"
   ]
  },
  {
   "id": "F-032",
   "kind": "fact",
   "claim": "Early ceiling test on the public dev set: even the gold passage leaves the 4B below a no-web frontier model.",
   "value": "raw: closed-book 0.487; full engine 0.507; with the gold passage 0.606; Claude Haiku without web 0.653",
   "ci": "n 57",
   "n": "57",
   "seeds": "1",
   "set": "v0.1 (EN)",
   "set_status": "public dev",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-04",
   "source_public": null,
   "caveat": "Raw ratios only; floor-adjusted values were not computed.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "canon:gold-oracle-v01",
    "F-system-75"
   ]
  },
  {
   "id": "F-033",
   "kind": "fact",
   "claim": "The first deep-tier test was the first lever to beat noise.",
   "value": "deep35 vs 4B: 0.565 vs 0.505, Δ +0.060 (+0.021…+0.101) raw",
   "ci": "paired 95 % bootstrap",
   "n": "57",
   "seeds": "2",
   "set": "v0.1",
   "set_status": "public dev",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "Raw ratio on the public dev set.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "canon:deep-first",
    "F-chron-early-28"
   ]
  },
  {
   "id": "F-034",
   "kind": "fact",
   "claim": "Mixture-of-experts models in the 25–35B class form one closed-book plateau, well above the 4B.",
   "value": "closed-book raw: Gemma 4 26B-A4B 0.613; Qwen3.5-35B-A3B 0.603; Qwen3.6-35B-A3B 0.596; vs 4B 0.487",
   "ci": "1 seed; n 57",
   "n": "57",
   "seeds": "1",
   "set": "v0.1",
   "set_status": "public dev",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-07",
   "source_public": "boar-research:reports/CLOSED.md:36; boar-research:reports/CLOSED.md:37; boar-research:reports/CLOSED.md:36-37; boar-research:reports/CLOSED.md#R17-R18",
   "caveat": "Screen differences between the MoE models are not detected at MDE 0.055.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "canon:moe-plateau",
    "F-results-57",
    "F-system-56"
   ]
  },
  {
   "id": "F-035",
   "kind": "fact",
   "claim": "Projected phone time for the recommended deep configuration (never measured on a phone).",
   "value": "≈ 1.9–2.8 min per answer at thinking budget 512 on a 12 GB phone CPU (third-party public decode rates of 4–6 tok/s + BOAR arithmetic); the 2026-10-09 estimate of 3.5–5 min was for budget 1024",
   "ci": "projection",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": "boar-research:reports/DEEP-RESEARCH-2026-10-12.md:8",
   "caveat": "Third-party rates, not a BOAR measurement.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "canon:deep-phone-time"
   ]
  },
  {
   "id": "F-036",
   "kind": "fact",
   "claim": "An expert-streaming engine runs the 35B model in under 3 GiB of RAM on a desktop CPU with judged-equivalent answers.",
   "value": "judged 0.625 vs llama-server 0.634, paired −0.010 (−0.042…+0.018); RSS 2.75–2.81 GiB; 14–17 tok/s on the Mac mini CPU; 10/10 greedy-identical vs the fork base",
   "ci": "paired 95 % bootstrap",
   "n": "57 (R450 contexts)",
   "seeds": "1",
   "set": "v0.1",
   "set_status": "public dev",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "Desktop only, experts read from a fast SSD and page cache; phone flash speed is unmeasured (analytic bound ≈ 4–6 tok/s with compute).",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "canon:les",
    "F-results-65",
    "F-results-66",
    "F-system-52",
    "F-chron-late-74"
   ]
  },
  {
   "id": "F-037",
   "kind": "fact",
   "claim": "The on-device MoE engine (bmoe) is not token-identical to the desktop server but is judged equivalent on identical contexts.",
   "value": "0.623 vs 0.634, Δ −0.012 (−0.045…+0.020); greedy token-identical 0/10",
   "ci": "paired 95 % bootstrap",
   "n": "57",
   "seeds": "1",
   "set": "v0.1",
   "set_status": "public dev",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-07",
   "source_public": "boar-research:reports/CLOSED.md:35; boar-research:reports/CLOSED.md#R16",
   "caveat": "An equivalence, not a lever: lets desktop deep-tier numbers transfer to a bmoe device plan with this caveat.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "canon:bmoe-parity",
    "F-results-64",
    "F-system-46",
    "F-chron-early-89"
   ]
  },
  {
   "id": "F-038",
   "kind": "fact",
   "claim": "Every touch of a held-out set was logged; both held-out sets are now development data.",
   "value": "v0.2: 12 full generation runs, 10 full Opus passes → burnt, dev (ruling 2026-10-06); V-para: 19 full runs + 12 slices → held-out for the 4B claim runs only, dev-style after 2026-10-06 (ruling 2026-10-09)",
   "ci": null,
   "n": "110 / 58 (51 judged + 7 venue)",
   "seeds": null,
   "set": "v0.2; V-para",
   "set_status": "burnt / dev-style",
   "judge": null,
   "date": "2026-10-09",
   "source_public": "boar-research:reports/SET-STATUS.md:14; boar-research:reports/SET-STATUS.md:15",
   "caveat": null,
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "canon:set-exposure"
   ]
  },
  {
   "id": "F-039",
   "kind": "fact",
   "claim": "Before the 2026-10-06 audit, a week of 4B improvements did not move the held-out score beyond noise.",
   "value": "v0.2 raw: main 0.506 → cut 6 0.512, two seeds each, +0.006 (noise)",
   "ci": null,
   "n": "110",
   "seeds": "2",
   "set": "v0.2",
   "set_status": "held-out at run time (burnt → dev 2026-10-06)",
   "judge": "Claude Opus (judge.v1, both orders)",
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "Do not compare 0.519 (one seed, s42) with the two-seed 0.506; the audit's \"+0.007\" and QUEUE:703 use other pairings.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "canon:heldout-move"
   ]
  },
  {
   "id": "F-040",
   "kind": "fact",
   "claim": "Deadlines: the bounty page (swept 2026-10-13) gives a final decision by 2026-10-31, and the issuer has said there is no fixed deadline; 2026-10-20 was the team's own closing-sprint deadline.",
   "value": "2026-10-31 (bounty page); 2026-10-20 (closing sprint)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "UNKNOWN whether the bounty itself moved to 10-20; call 10-20 the team's closing-sprint deadline unless the author confirms.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "canon:deadline"
   ]
  },
  {
   "id": "F-041",
   "kind": "fact",
   "claim": "By 2026-10-13 the ledger audit had verified 159 experiment records from 101 arm folders.",
   "value": "159 arm records (closed 99, lever-adopted 20, measurement 10, tooling 11, not-decided 9, parked 4, running 1, other 5)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": null,
   "caveat": "Another document counts 154; QUEUE:1089 (ledger audit v2) is the cited count.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "canon:arm-count"
   ]
  },
  {
   "id": "F-042",
   "kind": "fact",
   "claim": "Frozen pass/fail criteria were re-checked against their sha256 fingerprints.",
   "value": "65/65 criterion hashes verified (2026-10-10); an outside critic found 16 of 17 matching (2026-10-16), and 19 of 20 in a spot-check (1 hashing-convention mismatch, 1 criterion without a sidecar)",
   "ci": null,
   "n": "65 / 17 / 20",
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "Some early freezes are attested by file mtime only.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "canon:hash-check"
   ]
  },
  {
   "id": "F-043",
   "kind": "fact",
   "claim": "Negative results are kept in a public registry, each with its power and what would reopen it.",
   "value": "76 public rows (R1–R76) on 2026-10-14; 98 rows in the private registry by 2026-10-16",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": "boar-research:reports/CLOSED.md:6",
   "caveat": "Rows R77–R98 are not yet public and need a scrub (they mention internal items).",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "canon:closed-registry"
   ]
  },
  {
   "id": "F-044",
   "kind": "fact",
   "claim": "The floor-adjusted scale was pre-registered as a companion metric before any held-out result; it was promoted to headline after the audit.",
   "value": "2026-10-02: plan lists floor-adjusted as companion (plan sha 375ff004…); 2026-10-06: audit decision D1 makes it the headline with win/tie/loss; 2026-10-07: applied in claim draft v3.1",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": "boar-bench:bench/v0.3/plan/analysis-plan.md:22-25",
   "caveat": "The headline promotion came after the first held-out results; all three scales are always printed.",
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "canon:headline-switch"
   ]
  },
  {
   "id": "F-045",
   "kind": "fact",
   "claim": "How the win-rate condition changed: secondary metric in the plan, briefly part of the bar in arm reports, then \"reported, never gated\".",
   "value": "2026-10-02 plan: W/T/L secondary; 2026-10-08/09 arm reports read the bar as \"floor-adjusted CI low > 0.50 AND win-rate > 0\"; 2026-10-10 BAR-TABLE: win-rate reported, never gated",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": "boar-bench:bench/v0.3/plan/analysis-plan.md:30; boar-research:reports/BAR-TABLE.md:7",
   "caveat": null,
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "canon:winrate-conjunct"
   ]
  },
  {
   "id": "F-046",
   "kind": "fact",
   "claim": "The cut 6 freeze was relaxed once: cut 7 added the Wikidata fact table, which changed 8 public dev answers.",
   "value": "2026-10-06 rule: only routing/safety fixes with v0.1 answers unchanged; 2026-10-07: cut 7 = cut 6 + FACTS r6 changed 8 v0.1 answers (0 fires on the held-out sets)",
   "ci": null,
   "n": "8 of 57 v0.1 answers",
   "seeds": null,
   "set": "v0.1",
   "set_status": "public dev",
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CANONICAL (two-source check in verify pass)",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "canon:cut6-freeze"
   ]
  },
  {
   "id": "F-047",
   "kind": "event",
   "claim": "Public post (by Vitalik Buterin) asks for an offline phone research tool \">50 % as good as internet + frontier AI\"; bounty #31 is later opened on poidh, inspired by it",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-17",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "E-chron-early-L31"
   ]
  },
  {
   "id": "F-048",
   "kind": "fact",
   "claim": "Upstream app's own speed table (hot phone).",
   "value": "LFM2.5-8B-A1B 14.8 tok/s 5.2 GB; Qwen2.5-1.5B 11.4, 3.1 GB; Phi-3.5-mini 4.0; Qwen2.5-7B 2.7",
   "ci": "17 questions",
   "n": null,
   "seeds": null,
   "set": "maintainer's set",
   "set_status": null,
   "judge": null,
   "date": "2026-09-24",
   "source_public": null,
   "caveat": "not the research team's measurement",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-system-88"
   ]
  },
  {
   "id": "F-049",
   "kind": "event",
   "claim": "Upstream app README benchmark on a hot Xiaomi (Dimensity 8300): LFM2.5-8B-A1B 14.8 tok/s, Qwen2.5-1.5B 11.4 (maintainer's own measurement, 17 questions)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-24",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "T-system-L23"
   ]
  },
  {
   "id": "F-050",
   "kind": "decision",
   "claim": "Thesis: \"quality per second + blind evaluation against a frontier model\"; stay on the existing cross-platform app; plan a no-network Android submission — why: the existing claim had solid engineering but no measured research quality — outcome: led to the 09-29 research pipeline and 10-01 harness",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-26",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "D-chron-early-01"
   ]
  },
  {
   "id": "F-051",
   "kind": "event",
   "claim": "Author does a read-only review of the app maintainer's existing bounty claim (#124): \"solid engineering, weak research\"; default Qwen2.5-1.5B + ~1,300 Wikipedia paragraphs; biggest lever named = large streamed MoE + full Wikipedia",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-26",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "E-chron-early-L32"
   ]
  },
  {
   "id": "F-052",
   "kind": "event",
   "claim": "Author joins the app maintainer as co-builder; thesis \"quality per second + blind eval vs frontier\"; official Android submission planned as a no-INTERNET variant",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-26",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "E-chron-early-L33"
   ]
  },
  {
   "id": "F-053",
   "kind": "fact",
   "claim": "Claim kit packs.",
   "value": "boar-crypto 36.1 MB (3,141 articles), boar-wikivoyage-en 351 MB (34,002 articles)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-27",
   "source_public": "boar-research:packs/README.md:8-9",
   "caveat": "private lines are 6-7; 8-9 is the public copy",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-system-62"
   ]
  },
  {
   "id": "F-054",
   "kind": "event",
   "claim": "First knowledge kit packs built: boar-crypto (36.1 MB), boar-wikivoyage-en (351 MB), boar-preparedness",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-27",
   "source_public": "boar-research:packs/README.md:8-9,11",
   "caveat": "the cited l.8-11 are the public copy's line numbers; the private file has crypto/wikivoyage at l.6-7",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "T-system-L24"
   ]
  },
  {
   "id": "F-055",
   "kind": "event",
   "claim": "Overnight agent-run UI and layout-performance review of the app; heavy work routed to the Mac mini queue",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-28",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "E-process-L23"
   ]
  },
  {
   "id": "F-056",
   "kind": "event",
   "claim": "Research-pipeline kickoff: first written diagnosis (prefill dominates time-to-first-token; corpus coverage limits quality; 4B never run on a real Android)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-29",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "E-chron-early-L34"
   ]
  },
  {
   "id": "F-057",
   "kind": "event",
   "claim": "Overnight on-device tuning on an older 61-question bounty-style set (v2): score 0.26 → 0.48 with the 1.5B + quick mode + packs; vegan-venue test 5 % → 85 %",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-29",
   "date_text": "2026-09-29/30",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "E-chron-early-L35"
   ]
  },
  {
   "id": "F-058",
   "kind": "fact",
   "claim": "Earlier MoE probe: Qwen3.6-35B took 40–58 s to first token; CPU prefill was the limit",
   "value": "40–58 s",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-29",
   "source_public": null,
   "caveat": "pre-harness desktop/phone mix",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-chron-early-06"
   ]
  },
  {
   "id": "F-059",
   "kind": "decision",
   "claim": "References = Claude Opus + web search, frozen once, hashed; never regenerated (why: a fixed denominator makes runs comparable)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-30",
   "date_text": "2026-09-30 → 10-02",
   "source_public": "boar-bench:README.md:48-50; boar-bench:TRACE-FORMAT.md:79",
   "caveat": "outcome: same refs used for every v0.1 number; regenerating = not comparable",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "D-method-04"
   ]
  },
  {
   "id": "F-060",
   "kind": "event",
   "claim": "First \"in-RAM MoE\" experiments: Ling-3.0-tiny and Maple-Preview 20B-A1B vs Qwen3-4B (desktop, app contexts)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-30",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "E-chron-early-L36"
   ]
  },
  {
   "id": "F-061",
   "kind": "event",
   "claim": "Frozen frontier-plus-web reference answers for the public v0.1 set start being generated (09-30 → 10-02)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-30",
   "source_public": "boar-bench:README.md:50",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "E-chron-early-L37"
   ]
  },
  {
   "id": "F-062",
   "kind": "event",
   "claim": "Author decides 10 open items after an overnight eval run (early metric, iPhone 13; not comparable to later ratios)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-30",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "E-process-L24"
   ]
  },
  {
   "id": "F-063",
   "kind": "fact",
   "claim": "Before the harness, an overnight on-phone session lifted the old 61-question bounty-style score from 0.26 to 0.48",
   "value": "0.26 → 0.48",
   "ci": "CI 0.42–0.56, replicated in 2 runs",
   "n": "2",
   "seeds": null,
   "set": "old \"v2\" 61 items, BOAR's own dev set",
   "set_status": "dev",
   "judge": "Claude judge, both orders vs Opus+web",
   "date": "2026-09-30",
   "source_public": null,
   "caveat": "different instrument, model (1.5B), phone and set; NOT comparable to later v0.1/v0.2/V-para numbers",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-chron-early-04"
   ]
  },
  {
   "id": "F-064",
   "kind": "fact",
   "claim": "The same session took the vegan-venue test from 5 % to 85 % with offline places packs",
   "value": "5 % → 85 %",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "old v2 venue items",
   "set_status": null,
   "judge": "OSM objective check",
   "date": "2026-09-30",
   "source_public": null,
   "caveat": "pre-harness",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-chron-early-05"
   ]
  },
  {
   "id": "F-065",
   "kind": "fact",
   "claim": "An in-RAM sparse MoE (Ling-3.0-tiny 7.9B/1.3B active) read about as well as Qwen3-4B at ~2× speed on desktop",
   "value": "correct answers/min 6.0 vs 3.1",
   "ci": "56 answers/model",
   "n": null,
   "seeds": null,
   "set": "52-item dev set (precheck)",
   "set_status": "dev",
   "judge": "precheck v2 (not judge)",
   "date": "2026-09-30",
   "source_public": null,
   "caveat": "precheck, not judged ratio",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-chron-early-07"
   ]
  },
  {
   "id": "F-066",
   "kind": "fact",
   "claim": "Maple-Preview 20B-A1B (ternary) tied Qwen3-4B on 38 factual items with fewer confident-wrong",
   "value": "25/38 vs 26/38; CW 4 vs 7",
   "ci": "p = 1.00 paired",
   "n": null,
   "seeds": null,
   "set": "factual 38",
   "set_status": null,
   "judge": "blind hand labels",
   "date": "2026-09-30",
   "source_public": null,
   "caveat": "later full-engine test failed (F-chron-early-27)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-chron-early-08"
   ]
  },
  {
   "id": "F-067",
   "kind": "fact",
   "claim": "References were generated by Claude Opus 5.5 with WebSearch+WebFetch, one call at a time, 5 s gap, 1 retry, prompt reference.v1.",
   "value": "60 refs",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.1",
   "set_status": null,
   "judge": null,
   "date": "2026-09-30",
   "date_text": "2026-09-30 → 10-02",
   "source_public": "boar-bench:bench/v0.1/DATASET_CARD.md:87-90",
   "caveat": "references are AI output, \"not ground truth\"",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-method-04"
   ]
  },
  {
   "id": "F-068",
   "kind": "event",
   "claim": "Frozen Opus+web reference answers generated for the 33 kept v0 items (prompt reference.v1, sha d5144343…)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-09-30",
   "date_text": "2026-09-30 22:28-22:54Z",
   "source_public": "boar-bench:bench/v0.1/DATASET_CARD.md:87",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "T-method-L27"
   ]
  },
  {
   "id": "F-069",
   "kind": "decision",
   "claim": "Restate deep35 memory: \"≤ 8 GiB\" was our own estimate, not a bounty rule; macOS footprint ≠ phone RSS — why: the author asked why 8 GB if the bounty says 12 GB — outcome: Linux 6,750 MiB (≈ 6.6 GiB) container test passed 10/10",
   "value": "container capped at 6,750 MiB (about 6.6 GiB)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "Verification note: Same unit fix as F-chron-early-30.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "D-chron-early-12"
   ]
  },
  {
   "id": "F-070",
   "kind": "fact",
   "claim": "The bounty prize on 10-02 was 1.139 ETH (was 1.098), decision by 2026-10-31",
   "value": "1.139 ETH",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "n/a",
   "set_status": null,
   "judge": "n/a",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "later 1.52675 ETH (QUEUE:883); issuer address in file: do not publish",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-chron-early-01"
   ]
  },
  {
   "id": "F-071",
   "kind": "fact",
   "claim": "The bounty's hardware envelope.",
   "value": "≤ 12 GB RAM, ≤ 50 GB total storage, offline, no Play Services, GrapheneOS-compatible",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "GrapheneOS not yet run on hardware",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-system-92"
   ]
  },
  {
   "id": "F-072",
   "kind": "decision",
   "claim": "Close the knowledge-pack front for the bounty after the big-pack gate failed replay (\"no fourth pack iteration\") — why: kill rule K1 pre-registered in the strategy memo — outcome: packs reopened later only for the deep tier and as products",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 ~12:40",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "D-chron-early-08"
   ]
  },
  {
   "id": "F-073",
   "kind": "event",
   "claim": "Big-pack relevance gate fails replay → knowledge front closed for the bounty (kill rule K1)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 ~12:40",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "E-chron-early-L51"
   ]
  },
  {
   "id": "F-074",
   "kind": "decision",
   "claim": "Run the deep tier (35B MoE); its ratio counts only reported separately from the default, with its latency — why: first candidate lever with a prior above noise; bounty hint about large MoE — outcome: deep35 +0.060 (first lever above MDE)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 ~12:55",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "D-chron-early-09"
   ]
  },
  {
   "id": "F-075",
   "kind": "decision",
   "claim": "Single goal = win the bounty; English only; no more work inside the maintainer's app; build a dedicated measurement app — why: focus — outcome: PT/ES lines parked; BOAR Bench built",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "D-chron-early-18"
   ]
  },
  {
   "id": "F-076",
   "kind": "decision",
   "claim": "Single goal = the bounty bar; other quality work after the bounty (why: deadline + focus)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": "outcome: lanes re-assigned (model tier, health minimum, speed, demo device)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "D-system-02"
   ]
  },
  {
   "id": "F-077",
   "kind": "event",
   "claim": "Direction change: single goal = win the bounty",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "E-chron-early-L68"
   ]
  },
  {
   "id": "F-078",
   "kind": "event",
   "claim": "V-para blind set frozen (58 items, bounty-style paraphrases)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "E-chron-early-L70"
   ]
  },
  {
   "id": "F-079",
   "kind": "event",
   "claim": "Direction change: single goal = win the bounty; lanes re-tasked",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "E-process-L36"
   ]
  },
  {
   "id": "F-080",
   "kind": "fact",
   "claim": "A travel-distance calculator fixed one bounty-style item (Paris–Rome)",
   "value": "0.29 → 0.94",
   "ci": "1 item",
   "n": null,
   "seeds": null,
   "set": "V-exact (public quotes)",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-03",
   "source_public": null,
   "caveat": "built on that item; the V-exact 0.39 → 0.51 jump is this one item (PIPELINE-AUDIT:38)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-chron-early-50"
   ]
  },
  {
   "id": "F-081",
   "kind": "fact",
   "claim": "V-exact (5 judged bounty-quote items) on cut 2",
   "value": "0.510",
   "ci": "plain CI 0.303–0.739, n=5",
   "n": "5",
   "seeds": null,
   "set": "V-exact (public, dev)",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-03",
   "source_public": null,
   "caveat": "n=5; do not headline",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-chron-early-51"
   ]
  },
  {
   "id": "F-082",
   "kind": "event",
   "claim": "Direction change: single goal = win bounty #31 (>50 %)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "T-results-L25"
   ]
  },
  {
   "id": "F-083",
   "kind": "event",
   "claim": "Direction change: single goal = the bounty (>50%)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "T-system-L29"
   ]
  },
  {
   "id": "F-084",
   "kind": "decision",
   "claim": "Park the web-only knowledge pack for the bounty (public-figure blog pack and a research-forum pack were the alternatives).",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "source_public": null,
   "caveat": "outcome: later replaced by the open-corpus Knowledge Pack program (10-12)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "D-system-04"
   ]
  },
  {
   "id": "F-085",
   "kind": "event",
   "claim": "Knowledge-pack sizing: web-only pack parked for the bounty",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "T-system-L35"
   ]
  },
  {
   "id": "F-086",
   "kind": "fact",
   "claim": "First real-phone smoke (volunteer's 8 GB phone, below the bounty's 12 GB class): same engine tree hash as desktop, tiers 8/8 identical, model answers 100–147 s",
   "value": "TTFT 72–100 s; decode 2.0–2.9 tok/s",
   "ci": "9 items",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "n/a",
   "date": "2026-10-05",
   "source_public": null,
   "caveat": "device model: see sensitivity flags",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-chron-early-60"
   ]
  },
  {
   "id": "F-087",
   "kind": "decision",
   "claim": "Verdict labels LEVER / SMALL / NO / \"not decided at MDE\"; adopt only after two-seed confirmation on V-para — why: Stop over-reading noise — outcome: Adopted levers after 10-08: 0 on the bounty set",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "D-process-16"
   ]
  },
  {
   "id": "F-088",
   "kind": "fact",
   "claim": "Formal raw rule (CI low > 0.50) met only by the deep tier with thinking on both dev sets; floor-adjusted target met by none",
   "value": "V-para CI low 0.513 (seed noise); v0.2 0.569",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "V-para + v0.2 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-11",
   "date_text": "10-11",
   "source_public": null,
   "caveat": "the 0.50 floor-adj threshold is the program's own, not the bounty's",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-chron-late-06"
   ]
  },
  {
   "id": "F-089",
   "kind": "fact",
   "claim": "V-para is not pack-matched: few bounty-style items have their fact in any pack",
   "value": "6/51 installed packs; 10/51 with every pack on disk",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "V-para dev-style",
   "set_status": "dev-style",
   "judge": "jev",
   "date": "2026-10-12",
   "date_text": "10-12",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-chron-late-20"
   ]
  },
  {
   "id": "F-090",
   "kind": "fact",
   "claim": "A quarter of the v0.2 gap is the health-strict canned fallback route",
   "value": "18 v0.2 items carry 0.24 of the loss at ratio 0.43; on V-para all 51 judged items are on the deep-thinking path (0 health-strict)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.2 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-12",
   "date_text": "2026-10-12 08:56",
   "source_public": null,
   "caveat": "product issue invisible on the bounty-style set | Verification note: Source says 0 of the 51 judged V-para items, not \"0/58\".",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-chron-late-12"
   ]
  },
  {
   "id": "F-091",
   "kind": "fact",
   "claim": "Wave-1 packs barely cover the bounty-style set",
   "value": "V-para absent-item coverage 1/34 (union with kit 5/34)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "V-para dev-style",
   "set_status": "dev-style",
   "judge": "jev",
   "date": "2026-10-12",
   "date_text": "2026-10-12 23:59",
   "source_public": null,
   "caveat": "packs are product value, not a bar lever",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-chron-late-25"
   ]
  },
  {
   "id": "F-092",
   "kind": "fact",
   "claim": "All of Wikipedia (Aug-2025 FineWiki + 2026-10-01 delta) holds the missing fact for few bounty-style absent items (WIKIFULL)",
   "value": "9/34 verified (MARGINAL, lower bound)",
   "ci": "15 shards + delta",
   "n": null,
   "seeds": null,
   "set": "V-para dev-style",
   "set_status": "dev-style",
   "judge": "jev",
   "date": "2026-10-13",
   "date_text": "2026-10-13 01:50",
   "source_public": "boar-research:reports/CLOSED.md#R45",
   "caveat": "on v0.2 (dev) 37–38/78 — the general set's missing facts are mostly in Wikipedia",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-chron-late-17"
   ]
  },
  {
   "id": "F-093",
   "kind": "fact",
   "claim": "Share of Opus spent on the bounty-shaped set in week 2",
   "value": "≈ 5 % (204 of ≈ 4,380)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "V-para",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "another page says ≈ 10 % (≈ 420)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-process-26"
   ]
  },
  {
   "id": "F-094",
   "kind": "fact",
   "claim": "Yield collapsed after the first week",
   "value": "15 adopted improvements, all 10-04 → 10-08; ~70 later arms, 0 adopted points on the bounty metric",
   "ci": "159 arms",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "internal PT-BR page; counts from the arm index",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-process-48"
   ]
  },
  {
   "id": "F-095",
   "kind": "fact",
   "claim": "X2b no harm on the bounty-shaped kit",
   "value": "v0.1+simplewiki +0.012 (−0.019…+0.046); kit-only runs byte-identical",
   "ci": "n 57",
   "n": "57",
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Haiku",
   "date": "2026-10-14",
   "source_public": "boar-research:arms/headpass-st/HEADPASS-ST.md:77",
   "caveat": "KIT-X2b closed at ceiling (R93)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-results-89"
   ]
  },
  {
   "id": "F-096",
   "kind": "fact",
   "claim": "X2b (neighbour chunks + budgeted dense candidates appended after the keep step).",
   "value": "fact chunk in context 102/180 (from 35); +0.066 (+0.043…+0.091) Opus, +0.056 Haiku; v0.1 no-harm +0.012 (−0.019…+0.046); +≈ 420 prompt tokens p50",
   "ci": "n 180, 1 seed",
   "n": "180,",
   "seeds": "1",
   "set": "180 self-tests (dev)",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-14",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:21",
   "caveat": "not a bounty-bar result; needs format-v3 packs",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "F-system-73"
   ]
  },
  {
   "id": "F-097",
   "kind": "null-result",
   "claim": "Crypto selection gate (tf2): crypto passages only with crypto cue → one bounty item −0.018 (NO-GO by frozen bar)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "N-chron-early-L301"
   ]
  },
  {
   "id": "F-098",
   "kind": "null-result",
   "claim": "WIKIFULL (Wikipedia as kit content): all 15 FineWiki shards + 2026-10-01 delta → 9/34 (a lower bound) absent items verified (bounty-style facts mostly not in Wikipedia text)",
   "value": "9/34 verified (lower bound, MARGINAL); about a quarter at most of bounty-style absent facts are in Wikipedia text",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "Verification note: \"9/34 (a lower bound)\" inverts the bound.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "question",
   "aliases": [
    "N-chron-late-L272"
   ]
  },
  {
   "id": "F-099",
   "kind": "decision",
   "claim": "Replace the 27 items borrowed from a competitor's question list with own items (bench v0.1) — why: do not copy competitors' test cases; clean benchmark — outcome: v0.1 = 60 public items (57 EN judged)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-chron-early-03"
   ]
  },
  {
   "id": "F-100",
   "kind": "decision",
   "claim": "Freeze every deciding criterion (sha256) before data; blind sets written by separate writers, never read by researchers — why: avoid post-hoc tuning — outcome: 65/65 criterion shas verify by 10-10",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "date_text": "2026-10-01 (night)",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-chron-early-05"
   ]
  },
  {
   "id": "F-101",
   "kind": "decision",
   "claim": "Remove all 27 items taken from a competing team's test cases; replace each with an original item in the same slot (why: no competitor test cases in the bench)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "source_public": "boar-bench:bench/v0.1/DATASET_CARD.md:3,22,100",
   "caveat": "outcome: v0.1 = 60 items, residue check = 0",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-method-03"
   ]
  },
  {
   "id": "F-102",
   "kind": "event",
   "claim": "ADR-001 written: one harness, five layers, seven axes never averaged, gate vs full modes, frozen criteria",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-chron-early-L39"
   ]
  },
  {
   "id": "F-103",
   "kind": "fact",
   "claim": "Run-to-run noise of identical configurations was Δ 0.031 (17/60 labels changed)",
   "value": "0.031",
   "ci": "60 items",
   "n": null,
   "seeds": null,
   "set": "v0 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-01",
   "source_public": null,
   "caveat": "motivated the MDE rule",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-early-10"
   ]
  },
  {
   "id": "F-104",
   "kind": "fact",
   "claim": "Run-to-run noise on the night of 2026-10-01 was Δ 0.031 with 17/60 labels changed.",
   "value": "0.031",
   "ci": "n 60",
   "n": "60",
   "seeds": null,
   "set": "v0.1",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-01",
   "source_public": null,
   "caveat": "pre-dates the stats rules",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-84"
   ]
  },
  {
   "id": "F-105",
   "kind": "fact",
   "claim": "v0.1 new items: word-Jaccard max 0.30 vs the removed competitor questions, max 0.29 vs other public items, 0.33 vs held-out.",
   "value": "0.30 / 0.29 / 0.33",
   "ci": "n 27",
   "n": "27",
   "seeds": null,
   "set": "v0.1",
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "source_public": "boar-bench:bench/v0.1/DATASET_CARD.md:26",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-91"
   ]
  },
  {
   "id": "F-106",
   "kind": "event",
   "claim": "ADR-001 proposes one evaluation harness: 5 layers, one trace schema, 7 axes never averaged, gate vs full modes",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L28"
   ]
  },
  {
   "id": "F-107",
   "kind": "event",
   "claim": "Run-to-run noise found as large as measured effects (Δ 0.031, 17/60 labels changed)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "date_text": "2026-10-01 (night)",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L29"
   ]
  },
  {
   "id": "F-108",
   "kind": "event",
   "claim": "27 items from a competing team's fork removed from the bench; replaced with original items of the same slot (bench v0.1)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "source_public": "boar-bench:bench/v0.1/DATASET_CARD.md:3,100",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L30"
   ]
  },
  {
   "id": "F-109",
   "kind": "event",
   "claim": "Rule: an effect is reported only if it exceeds the minimum detectable effect (MDE)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "date_text": "2026-10-01 23:00",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-chron-early-L42"
   ]
  },
  {
   "id": "F-110",
   "kind": "fact",
   "claim": "Validated run budget: MDE 0.038 at 2 runs × 60 items; slot-A bias +0.11 cancels only with both orders",
   "value": "MDE 0.038",
   "ci": "predicted |Δ| ≤ 0.04, observed 0.006",
   "n": null,
   "seeds": null,
   "set": "v0.1-era public 60",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-01",
   "date_text": "2026-10-01 23:55",
   "source_public": null,
   "caveat": "Verification note: Cited lines 54-58 do not hold this; the run-budget row is QUEUE:59 (validated 2026-10-01 23:55).",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-early-11"
   ]
  },
  {
   "id": "F-111",
   "kind": "decision",
   "claim": "Treat the bounty as qualitative: judged by a public confirmation, not a numeric benchmark — why: bounty text — outcome: private ratio kept as instrument; claim framing open",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-chron-early-10"
   ]
  },
  {
   "id": "F-112",
   "kind": "decision",
   "claim": "Hold v0.2 as held-out: develop on v0.1, confirm only on v0.2 — why: avoid tuning on the claim set — outcome: v0.2 later burnt (D-chron-early-22)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-chron-early-11"
   ]
  },
  {
   "id": "F-113",
   "kind": "decision",
   "claim": "\"Iterate fast first, rigor later\": sha freeze only when an idea goes to the full bench or held-out; funnel L0 → L1 → L2 → L3 — why: speed of iteration — outcome: funnel became standard (BOUNTY-BOARD rules)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 night",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-chron-early-14"
   ]
  },
  {
   "id": "F-114",
   "kind": "decision",
   "claim": "Freeze the analysis plan (metrics, CI method, seeds, wording rules) before items existed (why: prevents tuning metric to results)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": "boar-bench:bench/v0.3/plan/README.md:3",
   "caveat": "outcome: plan hashed (375ff004…); later changes only as dated addenda",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-method-06"
   ]
  },
  {
   "id": "F-115",
   "kind": "event",
   "claim": "Held-out v0.2 analysis plan frozen (combined sha fe100e91…); items frozen in a second commit",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-chron-early-L54"
   ]
  },
  {
   "id": "F-116",
   "kind": "event",
   "claim": "Bounty re-read: winning is a public confirmation, not a numeric benchmark",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-chron-early-L56"
   ]
  },
  {
   "id": "F-117",
   "kind": "event",
   "claim": "Anchor ladder on v0.1: frontier no-web ≥ frontier+web → bar is effectively \"≥ 50 % of a frontier model\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-chron-early-L58"
   ]
  },
  {
   "id": "F-118",
   "kind": "event",
   "claim": "Held-out baseline E3: v0.2 raw 0.506 (0.475–0.538), v0.1 not inflated",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-chron-early-L59"
   ]
  },
  {
   "id": "F-119",
   "kind": "event",
   "claim": "Author: \"iterate fast first, rigor later\" (sha only when an idea reaches the full/held-out bench); no fixed greeting rules",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 night",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-chron-early-L62"
   ]
  },
  {
   "id": "F-120",
   "kind": "event",
   "claim": "Second freeze d8154b6 puts v0.2 question + reference plaintext into the private bench repo's git history (latent exposure)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": "boar-research:reports/SET-STATUS.md:14",
   "caveat": "Verification note: Line refs: EXPOSURES row is :28 (not :31); public v0.2 row is SET-STATUS:14 (not :17).",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-process-L32"
   ]
  },
  {
   "id": "F-121",
   "kind": "event",
   "claim": "Direction: two tracks only (harness + pipeline technique); v0.2 = held-out, confirm only",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 evening",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-process-L33"
   ]
  },
  {
   "id": "F-122",
   "kind": "fact",
   "claim": "Held-out baseline (E3) on v0.2",
   "value": "raw 0.506",
   "ci": "(0.475–0.538), 2 seeds; half-width 0.028 on 167",
   "n": "167",
   "seeds": "2",
   "set": "v0.2 held-out (at the time)",
   "set_status": "held-out (at run time)",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "floor-adj not given here",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-early-32"
   ]
  },
  {
   "id": "F-123",
   "kind": "fact",
   "claim": "The public bench has 60 English questions; 57 are judged (3 venue-list items scored against OpenStreetMap instead).",
   "value": "60 / 57",
   "ci": "n 60",
   "n": "60",
   "seeds": null,
   "set": "v0.1, public dev",
   "set_status": "public dev",
   "judge": null,
   "date": "2026-10-02",
   "source_public": "boar-bench:TRACE-FORMAT.md:17",
   "caveat": "25 items quote/paraphrase a named public author's posts",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-01"
   ]
  },
  {
   "id": "F-124",
   "kind": "fact",
   "claim": "v0.1 composition: explanation 11, reasoning 11, comparison 7, crypto 7, honesty-refusal 6, health 4, synthesis 3, local-food 3, danger-situation 3, danger 2, travel-practical 2, travel-local 1.",
   "value": "12 categories",
   "ci": "n 60",
   "n": "60",
   "seeds": null,
   "set": "v0.1",
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": "boar-bench:bench/v0.1/DATASET_CARD.md:62",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-03"
   ]
  },
  {
   "id": "F-125",
   "kind": "fact",
   "claim": "The judge scores each answer 1-5 on three axes: correctness (a material wrong claim caps it at 2; an uncorrected false premise at 1), completeness, usefulness.",
   "value": "3 axes, 1-5",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "all sets",
   "set_status": null,
   "judge": "judge.v1",
   "date": "2026-10-02",
   "source_public": "boar-bench:scorers/vendor/eval-lib/judge-core.mjs:40-44",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-05"
   ]
  },
  {
   "id": "F-126",
   "kind": "fact",
   "claim": "Above-decline ratio = (raw − D) / (1 − D) with D fixed at 0.306.",
   "value": "formula",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "all",
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": "boar-bench:bench/v0.3/plan/analysis-plan.md:24",
   "caveat": "different scale from floor-adjusted; does not use the floor",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-12"
   ]
  },
  {
   "id": "F-127",
   "kind": "fact",
   "claim": "CIs are a stratified (by category) two-level bootstrap: resample items within stratum, then runs within item; 2,000 resamples; seed 11; percentile 95 %.",
   "value": "B 2000, seed 11",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "all",
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": "boar-bench:metrics/floor_metrics.py:61-75; boar-bench:bench/v0.3/plan/analysis-plan.md:18",
   "caveat": "does not include run-level noise beyond the runs available (see F-method-41)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-16"
   ]
  },
  {
   "id": "F-128",
   "kind": "fact",
   "claim": "Calibration ladder on v0.1: Claude models with no web vs the Opus+web reference — Fable 1.048, Opus 1.029, Sonnet 0.950, Haiku 4.5 0.653 (raw).",
   "value": "raw 1.048 / 1.029 / 0.950 / 0.653",
   "ci": "1 run × 57 each; e.g. Opus 1.007-1.053",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "the reference behaves like \"frontier model memory\", web adds little on these items",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-23"
   ]
  },
  {
   "id": "F-129",
   "kind": "fact",
   "claim": "v0.2's items and references were written by Claude-family subagents (two blind writers, 55 each); V-para by an Opus subagent.",
   "value": "Claude-family",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "v0.2, V-para",
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02/03",
   "source_public": null,
   "caveat": "exact writer model id not recorded",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-62"
   ]
  },
  {
   "id": "F-130",
   "kind": "fact",
   "claim": "v0.2 writers converged: 36 same-topic pairs out of 55 between two Claude writers, then 7 of 37 again among rewriters.",
   "value": "36/55; 7/37",
   "ci": null,
   "n": "110 (inferred from set)",
   "seeds": null,
   "set": "v0.2",
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "citation fix (l.44 is an instruction step)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-63"
   ]
  },
  {
   "id": "F-131",
   "kind": "fact",
   "claim": "The v0.2 plan's E3 check: at ≈ 110 items the CI half-width alone is ≈ 0.037.",
   "value": "± 0.037",
   "ci": "n 110",
   "n": "110",
   "seeds": null,
   "set": "v0.2",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": "boar-bench:bench/v0.3/plan/analysis-plan.md:55",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-65"
   ]
  },
  {
   "id": "F-132",
   "kind": "fact",
   "claim": "Stratification by category: the plan reports the stratified interval with the plain one beside it.",
   "value": "both",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": "boar-bench:bench/v0.3/plan/analysis-plan.md:18; boar-bench:metrics/strata.py:1",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-85"
   ]
  },
  {
   "id": "F-133",
   "kind": "event",
   "claim": "Calibration ladder: Claude models with no web vs the Opus+web reference on v0.1 (Opus no-web 1.029, Sonnet 0.950, Haiku 4.5 0.653)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L32"
   ]
  },
  {
   "id": "F-134",
   "kind": "event",
   "claim": "Bench v0.2 analysis plan frozen and hashed before any v0.2 item existed (raw ratio = primary; stratified two-level CI; seeds 42/43/44)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": "boar-bench:bench/v0.3/plan/analysis-plan.md:1-20",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L34"
   ]
  },
  {
   "id": "F-136",
   "kind": "event",
   "claim": "v0.2 plaintext committed into the private bench repo's git history (second freeze) — latent exposure",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "never publish the commit history; say 'leaked into a private repo history'",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L36"
   ]
  },
  {
   "id": "F-137",
   "kind": "event",
   "claim": "References for the 27 new v0.1 items generated (27 calls, 0 failures, ≈ 1.68 USD list)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 00:08-00:18Z",
   "source_public": "boar-bench:bench/v0.1/DATASET_CARD.md:88",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L31"
   ]
  },
  {
   "id": "F-138",
   "kind": "decision",
   "claim": "Move v0.2 plaintext out of researcher-readable paths after an exposure; treat the exposed item as seen — why: a trace scan printed one held-out item — outcome: git history still held the plaintext → later \"burnt\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": "Verification note: Same date fix as the timeline row: the exposure is dated 10-03, not 10-02 22:05.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-chron-early-15"
   ]
  },
  {
   "id": "F-139",
   "kind": "event",
   "claim": "Ship candidate on held-out: raw 0.503, neutral vs baseline",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-chron-early-L72"
   ]
  },
  {
   "id": "F-140",
   "kind": "event",
   "claim": "Computed-verdict family (4 variants) closed: verdict accuracy +23–28 pp does not survive the judged bench",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "date_text": "2026-10-04/05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-chron-early-L82"
   ]
  },
  {
   "id": "F-141",
   "kind": "fact",
   "claim": "Cut 2 on v0.1 EN over four extra seeds",
   "value": "≈ 0.533 (0.532/0.535/0.539/0.527)",
   "ci": "4 seeds",
   "n": null,
   "seeds": "4",
   "set": "v0.1 EN dev",
   "set_status": "dev",
   "judge": "Opus-med",
   "date": "2026-10-04",
   "source_public": null,
   "caveat": "judge.v1-med ≡ v1 (κ 1.000)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-early-53"
   ]
  },
  {
   "id": "F-142",
   "kind": "fact",
   "claim": "Computed verdicts raised verdict accuracy +23–28 pp on blind verdict sets but lost judged quality",
   "value": "CVS +22.9 pp; L1 Δ −0.005",
   "ci": "83 scored; 4 changed items",
   "n": null,
   "seeds": null,
   "set": "blind gate sets / v0.1 EN",
   "set_status": null,
   "judge": "key + Opus-med",
   "date": "2026-10-04",
   "source_public": null,
   "caveat": "\"verdict accuracy and judged quality diverge\"",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-early-57"
   ]
  },
  {
   "id": "F-143",
   "kind": "event",
   "claim": "Cut 2 extra seeds: v0.1 ratio 0.532/0.535/0.539/0.527",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-results-L28"
   ]
  },
  {
   "id": "F-144",
   "kind": "fact",
   "claim": "V-para anchors (no web): Haiku 0.588, Sonnet 0.961, Opus 1.044",
   "value": "0.588 (0.559–0.615)",
   "ci": "1 seed",
   "n": null,
   "seeds": "1",
   "set": "V-para",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-05",
   "source_public": "boar-research:reports/BAR-TABLE.md:44",
   "caveat": "Haiku floor-adj 0.48 (CI unknown)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-early-64"
   ]
  },
  {
   "id": "F-145",
   "kind": "fact",
   "claim": "On V-para (no-web anchors): Haiku 0.588 raw (0.559-0.615) / floor-adj 0.48 (CI unknown), Sonnet 0.961, Opus 1.044.",
   "value": "see value",
   "ci": "1 run, n 51",
   "n": "51",
   "seeds": null,
   "set": "V-para (held-out at the time)",
   "set_status": "held-out at run time",
   "judge": "Opus",
   "date": "2026-10-05",
   "date_text": "2026-10-05/06",
   "source_public": "boar-research:reports/BAR-TABLE.md:44",
   "caveat": "Opus no-web 1.044 here vs 1.029 on v0.1",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-25"
   ]
  },
  {
   "id": "F-146",
   "kind": "fact",
   "claim": "No-web anchors Sonnet / Opus / Fable on V-para",
   "value": "0.961 / 1.044 / 1.037 raw",
   "ci": "n 51; 1 run",
   "n": "51",
   "seeds": null,
   "set": "V-para",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-05",
   "source_public": null,
   "caveat": "Opus judging Opus-written refs; same family",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-results-10"
   ]
  },
  {
   "id": "F-147",
   "kind": "decision",
   "claim": "Two-level judge funnel: L1 cheap triage on a 20-item slice, paired vs a named control; L2 Opus for confirmation; claim numbers Opus only; never mix judges in one number (why: 12k-20k judge calls in 6 days)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "outcome: spend fell to ≈ 1.5k / 2 days",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-method-11"
   ]
  },
  {
   "id": "F-148",
   "kind": "event",
   "claim": "v0.2 declared burnt → dev",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": "boar-research:reports/SET-STATUS.md:14",
   "caveat": "Verification note: Public v0.2 row is SET-STATUS:14.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-process-L38"
   ]
  },
  {
   "id": "F-149",
   "kind": "fact",
   "claim": "Retrieval scorecard on v0.1: recall@5 0.333; 19/39 gold misses are coverage",
   "value": "0.333",
   "ci": "39 gold items",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "deterministic",
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "replay harness 8 s/run, no LLM",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-early-77"
   ]
  },
  {
   "id": "F-150",
   "kind": "fact",
   "claim": "Sonnet-low vs Opus: 54/57 same winner, r 0.93, κ 0.68, ratio −0.037; Sonnet-medium 55/57, −0.03.",
   "value": "54/57",
   "ci": "n 57",
   "n": "57",
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Sonnet vs Opus",
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "Haiku 4.5 was slower (21 s/call, retries) and not used",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-48"
   ]
  },
  {
   "id": "F-151",
   "kind": "fact",
   "claim": "Earlier MDE: 2 seeds × 57 items ≈ 0.035; a 24-item L1 slice could not see +0.03.",
   "value": "0.035",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.1",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "citation fix (l.21 is held-out hygiene)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-68"
   ]
  },
  {
   "id": "F-152",
   "kind": "fact",
   "claim": "Judge effort \"medium\" (judge.v1-med) agreed with judge.v1 at κ 1.0 (effort flag likely a no-op for Opus).",
   "value": "κ 1.0",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.1",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "citation fix (l.20 is Power)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-93"
   ]
  },
  {
   "id": "F-153",
   "kind": "fact",
   "claim": "Over-refusal on V-para",
   "value": "0 of 102 (0 %)",
   "ci": "2 seeds",
   "n": "51 (inferred from set)",
   "seeds": "2",
   "set": "V-para held-out",
   "set_status": "held-out at run time",
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-results-105"
   ]
  },
  {
   "id": "F-154",
   "kind": "fact",
   "claim": "Safety on the 10 v0.2 danger items",
   "value": "0 unsafe-rule hits, both seeds",
   "ci": "n 10",
   "n": "10",
   "seeds": null,
   "set": "v0.2 held-out at time",
   "set_status": "held-out at run time",
   "judge": "rule-based",
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-results-106"
   ]
  },
  {
   "id": "F-155",
   "kind": "event",
   "claim": "L1 triage protocol frozen: 20-item slice, paired vs named control, never an absolute L1 ratio; Sonnet-low as cheap judge",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": "boar-research:reports/RULE-CHANGES.md:8",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L40"
   ]
  },
  {
   "id": "F-156",
   "kind": "event",
   "claim": "v0.2 ruled burnt (plaintext in private bench repo history) → dev",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": "boar-research:reports/SET-STATUS.md:14",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-results-L35"
   ]
  },
  {
   "id": "F-157",
   "kind": "decision",
   "claim": "Cheap judge for triage (Sonnet-low, replaced by Haiku 5.5-low on 10-07); Opus only for confirmation and claim numbers — why: judge is the largest cost — outcome: judge calls fell from 12k–20k calls / 6 days to ≈ 1.5k / 2 days",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "date_text": "2026-10-06 21:25",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-chron-early-25"
   ]
  },
  {
   "id": "F-158",
   "kind": "event",
   "claim": "Cheaper judge triage adopted (Sonnet-low, later Haiku)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "date_text": "2026-10-06 21:25",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-chron-early-L96"
   ]
  },
  {
   "id": "F-159",
   "kind": "decision",
   "claim": "Shard the judge: N = 4 default for Haiku L1 and for Opus L2/claim (why: ×4.5 wall gain, 0 rate limits, same winners)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "date_text": "2026-10-07 / 10-08",
   "source_public": "boar-research:method/L1-PROTOCOL.md:12-13",
   "caveat": "outcome: in force; a stopped shard is resumed, never skipped",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-method-14"
   ]
  },
  {
   "id": "F-160",
   "kind": "event",
   "claim": "Haiku 5.5-low adopted as L1 judge; judge sharding N = 4",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-chron-early-L103"
   ]
  },
  {
   "id": "F-161",
   "kind": "fact",
   "claim": "Fine-tune data under one strict faithfulness rubric: teacher 16/20, + sentence rejection 11/20, human CLAPnq+HAGRID 6/20 examples with an unsupported claim",
   "value": "16 / 11 / 6 of 20",
   "ci": "20-example reviews",
   "n": null,
   "seeds": null,
   "set": "n/a",
   "set_status": null,
   "judge": "Opus reviewer",
   "date": "2026-10-07",
   "source_public": "boar-research:reports/CLOSED.md:27",
   "caveat": "bar ≤ 2/20 unreachable even for human data",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-early-84"
   ]
  },
  {
   "id": "F-162",
   "kind": "fact",
   "claim": "Local open judge (Gemma 4 26B) agreed on the winner 49–50/57 but was lenient by +0.11",
   "value": "κ 0.32 / 0.55",
   "ci": "228 calls",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "local open judge",
   "date": "2026-10-07",
   "source_public": "boar-research:README.md",
   "caveat": "ranking preserved; not for the number",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-early-88"
   ]
  },
  {
   "id": "F-164",
   "kind": "fact",
   "claim": "Local Gemma 4 26B judge: same winner as Opus 49/57 and 50/57, κ 0.32 / 0.55, ratio +0.110 / +0.112 above Opus, 41-43 s/call.",
   "value": "+0.11",
   "ci": "228 calls, 2 legs",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Gemma vs Opus",
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-46"
   ]
  },
  {
   "id": "F-165",
   "kind": "fact",
   "claim": "Sharding: Haiku sequential 694 s → N = 4 154 s (44.4 calls/min) → N = 8 145 s; 0 rate-limit errors; same winner 55/57.",
   "value": "×4.5 / ×4.8",
   "ci": "114 calls/level",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Haiku",
   "date": "2026-10-07",
   "source_public": "boar-research:method/L1-PROTOCOL.md:12",
   "caveat": "wall at N ≥ 8 is straggler-bound",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-49"
   ]
  },
  {
   "id": "F-166",
   "kind": "fact",
   "claim": "Position consistency (both orders agree) is 96.5-100 % for Opus and Haiku, 86-93 % for the local Gemma judge.",
   "value": "96.5-100 %",
   "ci": "n 57",
   "n": "57",
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "several",
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-94"
   ]
  },
  {
   "id": "F-167",
   "kind": "fact",
   "claim": "Parallel judge shards raised throughput",
   "value": "9.9 → 44.4 calls/min (Haiku, N = 4); Opus 21.1 calls/min; 0 rate-limit errors",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "date_text": "2026-10-07/08",
   "source_public": "boar-research:reports/RULE-CHANGES.md:10,12",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-process-33"
   ]
  },
  {
   "id": "F-168",
   "kind": "event",
   "claim": "PARJUDGE: judge sharded into 4-8 parallel processes, ×4.5-4.8 faster, 0 rate-limit errors",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "date_text": "2026-10-07 18:06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L43"
   ]
  },
  {
   "id": "F-169",
   "kind": "decision",
   "claim": "SLICE20 triage before any full deep/held-out arm — why: ~60 % generation saving on NO arms — outcome: later replaced by stratified random slices (10-10)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-08",
   "source_public": "boar-research:reports/RULE-CHANGES.md:13",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-chron-early-29"
   ]
  },
  {
   "id": "F-170",
   "kind": "fact",
   "claim": "Opus in 4 shards: 20-call probe, 21.1 calls/min vs ≈ 6/min sequential, 0 rate limits, same winner 10/10, Δratio −0.014.",
   "value": "21.1/min",
   "ci": "20 calls, 10 items",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-08",
   "source_public": "boar-research:method/L1-PROTOCOL.md:13",
   "caveat": "10 items only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-50"
   ]
  },
  {
   "id": "F-171",
   "kind": "event",
   "claim": "OPUSSHARD: Opus judge also allowed in N = 4 shards (20-call probe, same winner 10/10)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-08",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L45"
   ]
  },
  {
   "id": "F-172",
   "kind": "event",
   "claim": "Exposure ledger built; V-para status ruling",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-chron-early-L107"
   ]
  },
  {
   "id": "F-173",
   "kind": "fact",
   "claim": "Thinking budget curve on V-para (floor-adj).",
   "value": "0 → 0.33, 256 → 0.39, 512 → 0.44, 1024 → 0.42; 3072 no gain",
   "ci": "2 seeds (3072: 1 seed Haiku)",
   "n": "51 (inferred from set)",
   "seeds": "2",
   "set": "V-para (dev-style)",
   "set_status": "dev-style",
   "judge": "Opus",
   "date": "2026-10-09",
   "source_public": "boar-research:README.md:37-44",
   "caveat": "knee at 512",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-system-43"
   ]
  },
  {
   "id": "F-174",
   "kind": "event",
   "claim": "Exposure ledger backfilled from the log (one row per touch of a blind set)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L48"
   ]
  },
  {
   "id": "F-175",
   "kind": "event",
   "claim": "Exposure ledger backfilled (45 rows); V-para status found to disagree across docs → ruling",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "date_text": "2026-10-09 00:39",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-process-L43"
   ]
  },
  {
   "id": "F-176",
   "kind": "decision",
   "claim": "Decide direction on v0.2 (seed-stable), confirm on V-para with 2 seeds; adopt a lever only after two-seed V-para confirmation (why: v0.2 per-item seed r 0.90 vs V-para 0.5-0.7)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": "boar-research:reports/RULE-CHANGES.md:17,24",
   "caveat": "outcome: in force",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-method-19"
   ]
  },
  {
   "id": "F-177",
   "kind": "decision",
   "claim": "Every criterion carries a Power block (expected Δ with source, n, seeds, MDE, slice); launcher refuses without it (why: 0/32 criteria had stated an MDE)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": "boar-research:reports/RULE-CHANGES.md:20; boar-research:method/CRITERION-TEMPLATE.md:19-29",
   "caveat": "outcome: enforced by `arm.sh launch` rc 7",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-method-21"
   ]
  },
  {
   "id": "F-178",
   "kind": "decision",
   "claim": "Random stratified 20-id slices (seed 7) replace the \"20 hardest items\" slice (why: hardest-20 inflated deltas (+0.091 slice vs +0.026 full))",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 / 10-11",
   "source_public": "boar-research:method/L1-PROTOCOL.md:28",
   "caveat": "outcome: in force; slice only screens for futility",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-method-22"
   ]
  },
  {
   "id": "F-179",
   "kind": "decision",
   "claim": "Fail-closed judge: abort unless every id judged in both orders; scripts print \"n judged = X of n planned = Y\" — why: Review found errored traces silently dropped and shards always exit 0 — outcome: Live for every later number",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-process-11"
   ]
  },
  {
   "id": "F-180",
   "kind": "decision",
   "claim": "Power block mandatory (expected Δ with source, n, seeds, MDE); launch refuses without it — why: 0/32 criteria since 10-07 had stated an expected Δ or MDE — outcome: Enforced in the arm driver (rc 7)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-process-12"
   ]
  },
  {
   "id": "F-182",
   "kind": "fact",
   "claim": "The reference answers score ≈ 4.9/5 on average, so the ratio behaves like an absolute rubric mean rather than a head-to-head.",
   "value": "≈ 4.9",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "v0.2, V-para",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:88-91",
   "caveat": "public citation fix: per-dimension reference means (4.63-5.00) are at l.88-91, not l.66-69",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-15"
   ]
  },
  {
   "id": "F-183",
   "kind": "fact",
   "claim": "Correctness-only ratio of the deep configuration exceeds 0.50 on both sets (V-para floor-adj 0.660, v0.2 0.743); the shortfall sits in completeness and usefulness.",
   "value": "0.660 / 0.743",
   "ci": "V-para 0.608-0.712",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:85-91",
   "caveat": "dev data; a per-axis view, not the metric",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-30"
   ]
  },
  {
   "id": "F-184",
   "kind": "fact",
   "claim": "\"20 hardest\" slices inflated deltas: +0.091 slice vs +0.026 full (THINK4B); +0.089 vs +0.009 (THINK4B-V02).",
   "value": "inflation ×3-10",
   "ci": "n 20 vs 51/110",
   "n": "20",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Opus/Haiku",
   "date": "2026-10-10",
   "source_public": "boar-research:method/L1-PROTOCOL.md:28",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-44"
   ]
  },
  {
   "id": "F-185",
   "kind": "fact",
   "claim": "0 of 32 criteria frozen between 10-07 and 10-10 stated an expected Δ or MDE.",
   "value": "0/32",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": "boar-research:reports/RULE-CHANGES.md:20",
   "caveat": "fixed by the Power block",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-70"
   ]
  },
  {
   "id": "F-187",
   "kind": "fact",
   "claim": "Before the Power block, no criterion stated an expected effect",
   "value": "0/32 criteria since 10-07",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-process-56"
   ]
  },
  {
   "id": "F-188",
   "kind": "fact",
   "claim": "Formal raw rule (CI low > 0.50 on both sets) reading",
   "value": "met by T and H on dev data only",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para + v0.2 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:72-80",
   "caveat": "H's v0.2 leg is 1 seed; desktop only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-results-36"
   ]
  },
  {
   "id": "F-189",
   "kind": "fact",
   "claim": "Win-rate vs the reference is 0 for every configuration and seed on V-para and v0.2 under the Claude judge.",
   "value": "0",
   "ci": "all rows",
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para, v0.2",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:80",
   "caveat": "v0.2 ties 1–2 at most",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-results-38"
   ]
  },
  {
   "id": "F-190",
   "kind": "fact",
   "claim": "Share of items with usefulness ≥ 3/5",
   "value": "K V-para 2.9 %, D 10.8 %, H 33.3 %; K v0.2 34.5 %, H 58.2 %",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para / v0.2",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:93",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-results-41"
   ]
  },
  {
   "id": "F-191",
   "kind": "fact",
   "claim": "Hybrid R2→H on V-para",
   "value": "raw 0.465 / 0.467 (vs K +0.052)",
   "ci": "per seed; plain CI",
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "Opus recombined",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:45",
   "caveat": "p50 13 s",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-results-62"
   ]
  },
  {
   "id": "F-192",
   "kind": "event",
   "claim": "Near-miss rule after a Haiku triage stopped an arm Opus later rated +0.048",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": "boar-research:method/L1-PROTOCOL.md:15-25",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L49"
   ]
  },
  {
   "id": "F-193",
   "kind": "event",
   "claim": "Independent ML-process review (RND-REVIEW): CIs ~1.2× too narrow, hardest-20 slices inflate deltas, 0/32 criteria stated an MDE",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L51"
   ]
  },
  {
   "id": "F-194",
   "kind": "event",
   "claim": "Fail-closed judge pipeline: errored or missing items now stop the run or score at the floor instead of vanishing",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 21:16",
   "source_public": "boar-research:reports/RULE-CHANGES.md:19",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L52"
   ]
  },
  {
   "id": "F-195",
   "kind": "decision",
   "claim": "Every deciding run needs a frozen criterion with a Power block; slices = random20 files — why: stop under-powered closures — outcome: arm.sh refuses launch without the block",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 22:09",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-chron-late-03"
   ]
  },
  {
   "id": "F-196",
   "kind": "event",
   "claim": "Power block mandatory in every criterion; stratified random 20-item slices replace \"hardest 20\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 22:09",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-process-L47"
   ]
  },
  {
   "id": "F-197",
   "kind": "event",
   "claim": "Power block mandatory in every criterion; random stratified 20-id slices replace the \"20 hardest\" slice",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 22:09",
   "source_public": "boar-research:reports/RULE-CHANGES.md:20-21",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L53"
   ]
  },
  {
   "id": "F-198",
   "kind": "decision",
   "claim": "Negative results relabelled \"not detected at MDE x (n, seeds)\" when under-powered (why: \"closed\" overstated weak negatives)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "source_public": "boar-research:reports/RULE-CHANGES.md:25; boar-research:method/METHOD.md:9",
   "caveat": "outcome: CLOSED.md rows carry MDE and reopen condition",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-method-24"
   ]
  },
  {
   "id": "F-199",
   "kind": "fact",
   "claim": "A 20-item random slice has SE ≈ 0.04-0.05: it only screens for futility.",
   "value": "SE 0.04-0.05",
   "ci": "n 20",
   "n": "20",
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "source_public": "boar-research:method/L1-PROTOCOL.md:28",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-45"
   ]
  },
  {
   "id": "F-200",
   "kind": "fact",
   "claim": "Fail-closed judge: `--strict` exits before any call if a trace/reference is missing; `--floor-errored` scores a failed answer 1/1/1 instead of dropping it.",
   "value": "flags",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "source_public": "boar-bench:harness/judge.mjs:8-10; boar-bench:TRACE-FORMAT.md:23,59",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-76"
   ]
  },
  {
   "id": "F-201",
   "kind": "fact",
   "claim": "Speculative pre-generation before slice verdicts wasted mini time",
   "value": "≈ 5.2 mini-h never judged",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.2, V-para",
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-process-44"
   ]
  },
  {
   "id": "F-202",
   "kind": "event",
   "claim": "Verdict labels LEVER / SMALL / NO / \"not detected at MDE\"; honesty relabel of under-powered closures",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "source_public": "boar-research:reports/RULE-CHANGES.md:24-25",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L55"
   ]
  },
  {
   "id": "F-203",
   "kind": "event",
   "claim": "CLOSED.md registry built (36 rows); honesty relabel \"not detected at MDE x\" for under-powered negatives",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "source_public": "boar-research:reports/CLOSED.md:3-9; boar-research:reports/RULE-CHANGES.md:25",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-results-L56"
   ]
  },
  {
   "id": "F-204",
   "kind": "fact",
   "claim": "What the bar means in loss terms",
   "value": "V-para floor-adj 0.44 → 0.50 = +0.06 floor-adj ≈ +0.048 raw ≈ one-tenth of what the best config still loses",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "V-para",
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "10-12",
   "source_public": "boar-research:reports/DEEP-RESEARCH-2026-10-12.md",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-late-92"
   ]
  },
  {
   "id": "F-205",
   "kind": "fact",
   "claim": "Bar test power at n = 110: required point estimate ≈ 0.53 without run noise, ≈ 0.545 with one-seed run noise; a true 0.53 passes only ≈ 29-50 % of the time.",
   "value": "0.53 / 0.545",
   "ci": "n 110",
   "n": "110",
   "seeds": null,
   "set": "v0.3 plan",
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": "boar-bench:bench/v0.3/plan/power-statement.md:14-18",
   "caveat": "t-based, not bootstrap",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-66"
   ]
  },
  {
   "id": "F-207",
   "kind": "decision",
   "claim": "Close QUANTCEIL as \"not detected at MDE 0.03\" after V-para two seeds — why: seeds split +0.030 / −0.022 — outcome: model-knowledge levers on this hardware declared exhausted",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 07:55",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-chron-late-07"
   ]
  },
  {
   "id": "F-208",
   "kind": "fact",
   "claim": "3-bit quant (IQ3_XXS) vs 2-bit on V-para, two seeds (QUANTCEIL)",
   "value": "Δ +0.004 raw; floor-adj 0.441 vs 0.436",
   "ci": "(−0.024…+0.033); seeds +0.030 / −0.022",
   "n": null,
   "seeds": null,
   "set": "V-para dev-style",
   "set_status": "dev-style",
   "judge": "Opus",
   "date": "2026-10-12",
   "date_text": "2026-10-12 07:55",
   "source_public": null,
   "caveat": "on v0.2 the wrong-bucket repair was +0.087 but did not carry to V-para; decode ×0.7",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-late-13"
   ]
  },
  {
   "id": "F-209",
   "kind": "fact",
   "claim": "V-para: 58 items (51 judged + 7 venue), 19 full generation runs and 12 slices of 20.",
   "value": "58 / 19 / 12",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para dev-style",
   "set_status": "dev-style",
   "judge": null,
   "date": "2026-10-13",
   "source_public": "boar-research:reports/SET-STATUS.md:15",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-59"
   ]
  },
  {
   "id": "F-210",
   "kind": "fact",
   "claim": "Ledger audit of 159 arm records: 92 closed decisive, 26 descriptive, 19 adopted, 12 tooling; 12 HIGH / 97 MEDIUM / 307 LOW discrepancies.",
   "value": "159 arms",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-71"
   ]
  },
  {
   "id": "F-211",
   "kind": "fact",
   "claim": "The planned 40-pair blind human label check (seed 7, stratified) has not been done.",
   "value": "0/40",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": "boar-bench:bench/v0.3/plan/analysis-plan.md:45-49",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-90"
   ]
  },
  {
   "id": "F-212",
   "kind": "fact",
   "claim": "V-para use count",
   "value": "19 full runs + 12 slices of 20 + 4 anchor models + 11 venue passes",
   "ci": "n 51 (+7 venue)",
   "n": "51",
   "seeds": null,
   "set": "V-para: held-out for 4B claim runs; dev-style after 10-06",
   "set_status": "held-out (at run time)",
   "judge": "Opus",
   "date": "2026-10-13",
   "source_public": "boar-research:reports/SET-STATUS.md:15",
   "caveat": "deep config picked as best of ≥ 8 → winner's-curse ≈ 0.01-0.02 | Verification note: Public V-para row is SET-STATUS:15 (not :18).",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-process-64"
   ]
  },
  {
   "id": "F-213",
   "kind": "event",
   "claim": "Ledger audit of 159 arms: 12 HIGH / 97 MEDIUM discrepancies, 48 verdicts without a registry row, 10 unlogged exposures",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "T-method-L57"
   ]
  },
  {
   "id": "F-214",
   "kind": "fact",
   "claim": "Claim red-team found no fabricated or inflated number",
   "value": "26 criterion/patch hashes verified; 9 overclaiming sentences; 11 cross-file mismatches",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 21:03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-late-51"
   ]
  },
  {
   "id": "F-215",
   "kind": "fact",
   "claim": "Pipeline audit found 9 harness defects, none moving a reported number",
   "value": "9 defects",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 21:11",
   "source_public": null,
   "caveat": "judge resume could splice foreign rows; fixed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-late-50"
   ]
  },
  {
   "id": "F-216",
   "kind": "decision",
   "claim": "Claim draft v3.9 adopts the red-team findings (v0.2 \"formerly held-out\", V-para writer disclosed, verdict reuse disclosed) — why: red-team found no inflated number but 9 overclaiming sentences — outcome: 41 edits applied",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 22:08",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "D-chron-late-19"
   ]
  },
  {
   "id": "F-217",
   "kind": "fact",
   "claim": "The v0.2 held-out set had 110 items, 12 full generation runs and 10 full Opus judge passes before being ruled dev.",
   "value": "110 / 12 / 10",
   "ci": null,
   "n": "110 (inferred from set)",
   "seeds": null,
   "set": "v0.2 burnt",
   "set_status": "burnt → dev",
   "judge": "Opus",
   "date": "2026-10-14",
   "source_public": "boar-research:reports/SET-STATUS.md:14",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-58"
   ]
  },
  {
   "id": "F-218",
   "kind": "fact",
   "claim": "Of 9 private blind sets, only one (a 32-item Spanish set) is still clean.",
   "value": "1/9 clean",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "all blind sets",
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:reports/SET-STATUS.md:21",
   "caveat": "Spanish line parked; not usable for the English claim",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-60"
   ]
  },
  {
   "id": "F-219",
   "kind": "fact",
   "claim": "v0.2 use count",
   "value": "12 full generation runs + 10 full Opus passes",
   "ci": "n 110",
   "n": "110",
   "seeds": null,
   "set": "v0.2, burnt → dev",
   "set_status": "burnt → dev",
   "judge": "Opus",
   "date": "2026-10-14",
   "source_public": "boar-research:reports/SET-STATUS.md:14",
   "caveat": "Verification note: Line fix only; value confirmed.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-process-63"
   ]
  },
  {
   "id": "F-220",
   "kind": "fact",
   "claim": "V-para exposure count",
   "value": "19 full generation runs + 12 slices of 20",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para",
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:reports/SET-STATUS.md:15",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-results-112"
   ]
  },
  {
   "id": "F-221",
   "kind": "fact",
   "claim": "v0.2 exposure count",
   "value": "12 full generation runs; 10 full Opus judge passes",
   "ci": null,
   "n": "110 (inferred from set)",
   "seeds": null,
   "set": "v0.2",
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:reports/SET-STATUS.md:14",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-results-113"
   ]
  },
  {
   "id": "F-222",
   "kind": "fact",
   "claim": "X1 (append neighbour chunks of retrieved article, 0 LLM)",
   "value": "+0.033 (+0.016…+0.050) Opus; fact present 35 → 72",
   "ci": "n 180",
   "n": "180",
   "seeds": null,
   "set": "180 self-tests dev",
   "set_status": "dev",
   "judge": "Opus (beyond frozen plan)",
   "date": "2026-10-14",
   "source_public": "boar-research:arms/headpass-st/HEADPASS-ST.md:29,45",
   "caveat": "frozen label \"not decidable at MDE 0.03\" | verdict capped \"not decidable\": power line missed by 3 items",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-results-85",
    "F-chron-late-37"
   ]
  },
  {
   "id": "F-223",
   "kind": "event",
   "claim": "Open-source intent: research AND benchmark public; v0.1 reference-answer text made public",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 22:17–22:20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-chron-late-L71"
   ]
  },
  {
   "id": "F-225",
   "kind": "fact",
   "claim": "Public release size and review",
   "value": "bench 139 files / 2.26 MB; research 411 files / 5.53 MB; reviewer: 7 must-fix, 13 should-fix, 12 nits; 41/44 applied",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": "boar-research:README.md",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-process-71"
   ]
  },
  {
   "id": "F-226",
   "kind": "fact",
   "claim": "Claim section 0 length cap",
   "value": "≤ 1,100 words (raised from 900 for the judge-family paragraph)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-late-85"
   ]
  },
  {
   "id": "F-227",
   "kind": "fact",
   "claim": "Critique: most traced numbers re-derive",
   "value": "15/17 numbers, 16/17 criterion hashes match",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-late-86"
   ]
  },
  {
   "id": "F-228",
   "kind": "fact",
   "claim": "Judge-auth outage",
   "value": "judge passes impossible ≈ 06:4x → 09:0x on 10-16",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "no pass lost",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-chron-late-98"
   ]
  },
  {
   "id": "F-229",
   "kind": "fact",
   "claim": "Under-powered closures still labelled \"closed\"",
   "value": "≥ 6-7 rows with CI upper ≥ +0.03 at n ≈ 20",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "slices (dev)",
   "set_status": "dev",
   "judge": "Haiku/Opus",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "relabel to \"not detected at MDE 0.08\"",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-process-61"
   ]
  },
  {
   "id": "F-230",
   "kind": "event",
   "claim": "Cross-family judge (author via Codex) on existing V-para traces: Claude judge ≈ +0.10 stricter",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 19:01",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-process-L72"
   ]
  },
  {
   "id": "F-231",
   "kind": "fact",
   "claim": "The exposure ledger holds 126 dated rows (55 v0.2-only, 33 V-para-only, 19 both).",
   "value": "126 dated rows (55 v0.2-only, 33 V-para-only, 19 both)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "source_public": null,
   "caveat": "recount of rows starting '| 10-' in the Ledger section gives 126, not 128; the 55/33/19 split is confirmed",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-61"
   ]
  },
  {
   "id": "F-232",
   "kind": "fact",
   "claim": "The Claude judge is the stricter family by ≈ 0.10 raw on both sets",
   "value": "offset +0.08…+0.11 raw (+0.08/+0.115 V-para)",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para, v0.2",
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "source_public": null,
   "caveat": "ranking unchanged",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-results-53"
   ]
  },
  {
   "id": "F-233",
   "kind": "event",
   "claim": "Cross-family offset replicates on v0.2",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "date_text": "2026-10-17 01:09",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "E-process-L73"
   ]
  },
  {
   "id": "F-234",
   "kind": "fact",
   "claim": "The first A/B order per item is drawn with a seeded PRNG (mulberry32, seed 7); both orders always run.",
   "value": "seed 7",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "all",
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-bench:scorers/vendor/eval-lib/judge-core.mjs:22-38; boar-bench:TRACE-FORMAT.md:53",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-08"
   ]
  },
  {
   "id": "F-235",
   "kind": "fact",
   "claim": "Combining the two orders: same winner keeps it, disagreement becomes a tie; rubric scores are averaged.",
   "value": "rule",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "all",
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-bench:scorers/vendor/eval-lib/judge-core.mjs:93-98",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-09"
   ]
  },
  {
   "id": "F-236",
   "kind": "fact",
   "claim": "Raw ratio = Σ S / Σ R, where S and R are the app's and reference's mean rubric (mean of 3 axes, both orders averaged).",
   "value": "formula",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "all",
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-bench:metrics/floor_metrics.py:6,10; boar-bench:TRACE-FORMAT.md:67",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-10"
   ]
  },
  {
   "id": "F-237",
   "kind": "fact",
   "claim": "Floor-adjusted ratio = Σ (S − 1) / Σ (R − 1): quality above the rubric minimum of 1; 0.50 here means \"half as good as the reference\".",
   "value": "formula",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "all",
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-bench:metrics/floor_metrics.py:7,59; boar-bench:TRACE-FORMAT.md:68",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-11"
   ]
  },
  {
   "id": "F-238",
   "kind": "fact",
   "claim": "Strata merge danger-situation → danger and travel-local/practical → travel.",
   "value": "9 strata (v0.2)",
   "ci": null,
   "n": "110 (inferred from set)",
   "seeds": null,
   "set": "v0.2",
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-bench:metrics/paired_gap.py:9; boar-bench:bench/v0.3/plan/analysis-plan.md:9",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-17"
   ]
  },
  {
   "id": "F-239",
   "kind": "fact",
   "claim": "Paired differences between two systems use the same stratified two-level bootstrap on the three scales.",
   "value": "B 2000, seed 11",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "all",
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-bench:metrics/paired_gap.py:1-43",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-18"
   ]
  },
  {
   "id": "F-241",
   "kind": "fact",
   "claim": "Every judgment row records judgeModel, cliVersion, judgeVersion and the three input shas.",
   "value": "6 provenance fields",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-bench:TRACE-FORMAT.md:78",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-78"
   ]
  },
  {
   "id": "F-242",
   "kind": "fact",
   "claim": "Judge versions are tagged and never pooled: judge.v1 (Opus), judge.v1-low (cheap effort), judge.v1-med.",
   "value": "tags",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-bench:harness/judge.mjs:33",
   "caveat": "citation fix: the version tag is built at l.33",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-79"
   ]
  },
  {
   "id": "F-243",
   "kind": "fact",
   "claim": "Everything after the judge (traces → metrics) is deterministic given the judgment rows.",
   "value": "deterministic",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-bench:TRACE-FORMAT.md:80",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-80"
   ]
  },
  {
   "id": "F-244",
   "kind": "fact",
   "claim": "Expected public CI half-width at n = 57: ≈ ± 0.04 to 0.06.",
   "value": "± 0.04-0.06",
   "ci": "n 57",
   "n": "57",
   "seeds": null,
   "set": "v0.1",
   "set_status": null,
   "judge": "Opus",
   "date": null,
   "source_public": "boar-bench:README.md:27",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "F-method-95"
   ]
  },
  {
   "id": "F-245",
   "kind": "null-result",
   "claim": "Computed verdicts (CV, CVS, CVH, CVO): compute comparison verdicts → +23–28 pp verdict accuracy, judged quality ≤ 0 (family closed)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-chron-early-L306"
   ]
  },
  {
   "id": "F-246",
   "kind": "null-result",
   "claim": "VITALADD: one intro chunk added without displacement → Opus slice −0.011; full v0.2 L2 +0.001 (−0.012…+0.015) (lead sections rarely carry the rubric's specific fact)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-chron-late-L263"
   ]
  },
  {
   "id": "F-247",
   "kind": "null-result",
   "claim": "Sonnet-low as cheap judge: L1 triage judge 10-06",
   "value": "worked (54/57) but 14× more expensive than Haiku",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: replaced by Haiku 5.5-low",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-method-L212"
   ]
  },
  {
   "id": "F-248",
   "kind": "null-result",
   "claim": "Haiku 4.5 as judge: calibration run stopped at 46/114 calls",
   "value": "21 s/call, structured-output retries",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: slower and unverified",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-method-L213"
   ]
  },
  {
   "id": "F-249",
   "kind": "null-result",
   "claim": "\"20 hardest items\" L1 slice: signal20 slice for triage",
   "value": "inflated deltas (×3 to ×10 vs full set)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:method/L1-PROTOCOL.md:28",
   "caveat": "why closed: heterogeneity; replaced by random stratified slices",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-method-L216"
   ]
  },
  {
   "id": "F-250",
   "kind": "null-result",
   "claim": "Point-estimate promotion rules at L1: promote if Δ ≥ +0.03",
   "value": "a true +0.03 arm stopped ≈ 40-60 % of the time",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: replaced by futility rule + near-miss escalation",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-method-L217"
   ]
  },
  {
   "id": "F-251",
   "kind": "null-result",
   "claim": "N = 8 judge shards: parallel judging",
   "value": "only 9 s faster than N = 4 (straggler-bound)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:method/L1-PROTOCOL.md:12",
   "caveat": "why closed: N = 4 kept as default",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-method-L220"
   ]
  },
  {
   "id": "F-252",
   "kind": "null-result",
   "claim": "\"Seed noise\" threshold 0.03: rule for one-seed deep deltas",
   "value": "seeds differed by 0.039 with CI excluding 0",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: rule raised to ≥ 0.05 or two seeds",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-method-L224"
   ]
  },
  {
   "id": "F-253",
   "kind": "null-result",
   "claim": "Speculative full-set pre-generation: Generate full v0.2 + V-para before slice verdicts → ≈ 5.2 mini-h never judged (Both arms died at the slice; rule \"slice before full\")",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-process-L202"
   ]
  },
  {
   "id": "F-254",
   "kind": "null-result",
   "claim": "Same-winner as judge-validity statistic: \"Haiku same winner as Opus 54/57\" → uninformative (Reference wins ≈ 100 % of items)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-process-L213"
   ]
  },
  {
   "id": "F-255",
   "kind": "null-result",
   "claim": "R37 3-bit quant of deep model: 2 seeds V-para + v0.2",
   "value": "+0.004 / +0.008",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:61",
   "caveat": "why closed: not detected; ≥ +0.03 excluded",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-results-L264"
   ]
  },
  {
   "id": "F-256",
   "kind": "null-result",
   "claim": "R68 Vital Articles as additive quota: v0.2 110 Opus",
   "value": "+0.0012 (−0.012…+0.015)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:97",
   "caveat": "why closed: not detected at 0.025",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-results-L289"
   ]
  },
  {
   "id": "F-257",
   "kind": "null-result",
   "claim": "R84 Source-bound reading rule (lite): slice 20",
   "value": "−0.001 (−0.064…+0.063)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "R84 n 20, CI upper +0.063: 'not detected at MDE 0.08' (CRITIQUE-2026-10-16.md:59)",
   "verification": "OVERCLAIM",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-results-L305"
   ]
  },
  {
   "id": "F-258",
   "kind": "null-result",
   "claim": "R85 \"3–4 sentences\" style line (lite): slice 20",
   "value": "+0.016 (−0.035…+0.066)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "R85 n 20, CI upper +0.066: 'not detected at MDE 0.08' (CRITIQUE-2026-10-16.md:59)",
   "verification": "OVERCLAIM",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-results-L306"
   ]
  },
  {
   "id": "F-259",
   "kind": "null-result",
   "claim": "R87 Fact selection for lite: 0-call ceiling",
   "value": "≈ +0.014",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: ceiling below MDE",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-results-L308"
   ]
  },
  {
   "id": "F-260",
   "kind": "null-result",
   "claim": "R89 Health no-source answer from memory: 0 calls re-scope",
   "value": "ceiling +0.024; 0 V-para effect",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: closed by overlap; reverses never-from-memory health rule",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-results-L310"
   ]
  },
  {
   "id": "F-261",
   "kind": "null-result",
   "claim": "R94 Detector false fires on non-health v0.2: 0 calls",
   "value": "ceiling +0.0185",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: below MDE 0.02",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-results-L315"
   ]
  },
  {
   "id": "F-262",
   "kind": "null-result",
   "claim": "Computed verdict lines: CV/CVS/CVH/CVO",
   "value": "+23–28 pp verdict accuracy, but judged quality flat or negative",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: replacing prose loses quality",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "ruler",
   "aliases": [
    "N-system-L220"
   ]
  },
  {
   "id": "F-263",
   "kind": "fact",
   "claim": "D = 0.306 = mean judge rubric of 79 declined answers across 7 runs (≈ 1.50/5) divided by the mean reference rubric; v0.2-weighted 0.307.",
   "value": "0.306",
   "ci": "n 79 declines, 7 runs",
   "n": "79",
   "seeds": null,
   "set": "v0.1 night runs, dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-01",
   "source_public": "boar-bench:bench/v0.3/plan/analysis-plan.md:24",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-method-13"
   ]
  },
  {
   "id": "F-264",
   "kind": "decision",
   "claim": "Bar judged on raw ratio; raw + floor-removed always shown side by side — why: Author ruling — outcome: Superseded in emphasis by D-process-08",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "D-process-05"
   ]
  },
  {
   "id": "F-265",
   "kind": "fact",
   "claim": "Two-seed 4B baseline on v0.1",
   "value": "raw 0.512",
   "ci": "(0.461–0.567), 2 seeds, MDE 0.030",
   "n": null,
   "seeds": "2",
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "floor-removed ≈ 0.387 (QUEUE:174-177)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-chron-early-13"
   ]
  },
  {
   "id": "F-266",
   "kind": "fact",
   "claim": "iPhone-profile 1.5B baseline",
   "value": "raw 0.379",
   "ci": "(0.341–0.419), 2 seeds",
   "n": null,
   "seeds": "2",
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "floor-removed 0.219",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-chron-early-14"
   ]
  },
  {
   "id": "F-267",
   "kind": "fact",
   "claim": "The judge schema forces each rubric score ≥ 1, so an always-decline system scores raw ≈ 0.306",
   "value": "D = 0.306",
   "ci": "79 judged declines",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": "boar-research:reports/BAR-TABLE.md:14",
   "caveat": "defines the floor-adjusted scales",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-chron-early-20"
   ]
  },
  {
   "id": "F-268",
   "kind": "fact",
   "claim": "On the 4B, raw 0.512 = 0.387 floor-removed = 0.297 above always-decline",
   "value": "0.512 / 0.387 / 0.297",
   "ci": "2 seeds; floor-removed CI 0.319–0.448",
   "n": null,
   "seeds": "2",
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "three scales for one run; v0.1 dev, 2026-10-02; the near-identical 0.512 on v0.2 (F-002) is a different set.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-chron-early-21"
   ]
  },
  {
   "id": "F-269",
   "kind": "fact",
   "claim": "Anchor ladder: frontier models without web score ≥ 1.0; Haiku no-web 0.653; 4B 0.512; 1.5B 0.394; always-decline 0.306",
   "value": "1.048 / 1.029 / 0.950 / 0.653 / 0.512 / 0.394 / 0.306",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "Opus no-web > Opus+web by +0.029 (77 % ties): web adds nothing this judge sees",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-chron-early-31"
   ]
  },
  {
   "id": "F-270",
   "kind": "fact",
   "claim": "Cross-family rule in the plan: if a second-family judge differs by > 0.05 raw, the lower of the two is the headline; report κ between judges.",
   "value": "0.05",
   "ci": null,
   "n": "110 (inferred from set)",
   "seeds": null,
   "set": "v0.2/v0.3",
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": "boar-bench:bench/v0.3/plan/analysis-plan.md:39-44",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-method-89"
   ]
  },
  {
   "id": "F-271",
   "kind": "decision",
   "claim": "Bar on the RAW ratio; every publication shows raw + floor-removed + above-always-decline side by side — why: rubric floor gives a blank system ~0.20–0.31 — outcome: headline later switched to floor-adjusted (D-chron-early-21) while raw stays the formal rule",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 ~11:15",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "D-chron-early-07"
   ]
  },
  {
   "id": "F-272",
   "kind": "event",
   "claim": "Author's morning decisions: bar on the raw ratio, but always publish raw + floor-removed + above-decline side by side",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 ~11:15",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "E-chron-early-L50"
   ]
  },
  {
   "id": "F-273",
   "kind": "event",
   "claim": "Author rules the bar is on the raw ratio; every publication shows raw + floor-removed side by side",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 ~11:15",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "E-process-L30"
   ]
  },
  {
   "id": "F-274",
   "kind": "fact",
   "claim": "Ship candidate on held-out v0.2: quality-neutral vs baseline",
   "value": "raw 0.503",
   "ci": "(0.468–0.539), Δ −0.004 (−0.032…+0.026), 2 seeds",
   "n": null,
   "seeds": "2",
   "set": "v0.2 held-out (at the time)",
   "set_status": "held-out (at run time)",
   "judge": "Opus",
   "date": "2026-10-03",
   "source_public": null,
   "caveat": "above-min 0.376, above-decline 0.284",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-chron-early-44"
   ]
  },
  {
   "id": "F-275",
   "kind": "fact",
   "claim": "Claim runs on cut 3: held-out v0.2 raw 0.513 (\"consistent with, not shown above 0.50\"), floor-removed 0.389",
   "value": "0.513 / 0.389",
   "ci": "(0.479–0.550), 2 seeds; Δ vs ship +0.011 < MDE 0.018",
   "n": null,
   "seeds": "2",
   "set": "v0.2 held-out (at the time)",
   "set_status": "held-out (at run time)",
   "judge": "Opus",
   "date": "2026-10-05",
   "source_public": null,
   "caveat": "above-decline 0.299",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-chron-early-62"
   ]
  },
  {
   "id": "F-276",
   "kind": "fact",
   "claim": "4B claim build on V-para (two seeds): raw 0.413, floor-adj 0.261, above-decline 0.155, win/tie/loss 0/0/51 each seed.",
   "value": "0.413 / 0.261",
   "ci": "raw 0.384-0.445; floor 0.224-0.300; n 51; s42+s43",
   "n": "51",
   "seeds": null,
   "set": "V-para, held-out at run time",
   "set_status": "held-out at run time",
   "judge": "Opus judge.v1",
   "date": "2026-10-05",
   "date_text": "2026-10-05/06",
   "source_public": "boar-research:reports/BAR-TABLE.md:35",
   "caveat": "citation fix: l.31 is a header. These are the only held-out-grade numbers",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-method-26"
   ]
  },
  {
   "id": "F-277",
   "kind": "decision",
   "claim": "Lead every public number with floor-adjusted + win-rate, not raw (why: raw 0.50 is a rubric-floor artefact (always-decline ≈ 0.31 raw))",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "date_text": "2026-10-06 (audit decision D1) / 2026-10-07 (applied in claim draft v3.1)",
   "source_public": "boar-research:README.md:15-17; boar-bench:bench/v0.3/plan/analysis-plan.md:22-25",
   "caveat": "method notes date the switch 10-06, results notes 10-07; both are right for different steps. Also: the floor scale itself was pre-registered as a companion metric on 10-02, which answers the 'moving goalposts' attack",
   "verification": "CONFLICT",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "D-results-04"
   ]
  },
  {
   "id": "F-278",
   "kind": "fact",
   "claim": "Claim runs on cut 6 (two seeds): v0.2 raw 0.512, floor-removed 0.388, win 0/110; V-para unchanged 0.413",
   "value": "0.512 (0.478–0.549) / 0.388 (0.346–0.434)",
   "ci": "2 seeds",
   "n": null,
   "seeds": "2",
   "set": "v0.2 (held-out at the run; burnt → dev after)",
   "set_status": "held-out (at run time)",
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": "boar-research:reports/BAR-TABLE.md:53",
   "caveat": "Δ vs cut 3 −0.001; s42-only 0.519 / 0.397",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-chron-early-71"
   ]
  },
  {
   "id": "F-279",
   "kind": "fact",
   "claim": "V-para 4B claim numbers (two seeds), three scales",
   "value": "raw 0.413 (0.384–0.445), floor-adj 0.261 (0.224–0.300), above-decl 0.155, W/T/L 0/0/51",
   "ci": "2 seeds",
   "n": null,
   "seeds": "2",
   "set": "V-para held-out for this run",
   "set_status": "held-out (at run time)",
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": "boar-research:reports/BAR-TABLE.md:35",
   "caveat": "correctness ≤ 2 on 49 % (PIPELINE-AUDIT:14)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-chron-early-72"
   ]
  },
  {
   "id": "F-280",
   "kind": "fact",
   "claim": "4B claim build on v0.2 (two seeds): raw 0.512, floor-adj 0.388, above-decline 0.297, W/T/L 0/0/110 and 0/1/109.",
   "value": "0.512 / 0.388",
   "ci": "raw 0.478-0.549; floor 0.346-0.434; n 110; s42+s43",
   "n": "110",
   "seeds": null,
   "set": "v0.2, burnt → dev",
   "set_status": "burnt → dev",
   "judge": "Opus judge.v1",
   "date": "2026-10-06",
   "source_public": "boar-research:reports/BAR-TABLE.md:53",
   "caveat": "citation fix: l.52 is a separator; s42 alone 0.519 / 0.397",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-method-27"
   ]
  },
  {
   "id": "F-281",
   "kind": "event",
   "claim": "Pipeline audit: rubric floor gives an always-decline system ≈ 0.31 \"for free\"; decision D1 = floor-adjusted headline; D5 = treat v0.2 as burnt",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "citation fix: always-decline ≈ 0.31 is l.12, D1 l.53, D5 l.57 (cited 13/63/67 point at other rows)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "T-method-L39"
   ]
  },
  {
   "id": "F-282",
   "kind": "event",
   "claim": "Honest-metric rewrite: headline becomes floor-adjusted ratio + win/tie/loss, raw second",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "date_text": "2026-10-06 22:40",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "E-process-L39"
   ]
  },
  {
   "id": "F-283",
   "kind": "decision",
   "claim": "Local open-weights judge (Gemma 4 26B) allowed for triage/ranking only, never for the number (why: keeps ranking, +0.11 lenient, κ 0.32-0.55)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": "outcome: not used for headline numbers",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "D-method-13"
   ]
  },
  {
   "id": "F-284",
   "kind": "fact",
   "claim": "On correctness alone H is above 0.50 floor-adj; the gap is in completeness and usefulness",
   "value": "V-para correctness floor-adj 0.660 (0.608–0.712) vs completeness 0.301 / usefulness 0.341",
   "ci": "2 seeds",
   "n": "51 (inferred from set)",
   "seeds": "2",
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:89,95",
   "caveat": "\"correct but thin\" also describes an always-decline system on those dims; not a bar reading",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-results-39"
   ]
  },
  {
   "id": "F-285",
   "kind": "fact",
   "claim": "In-pack conditional quality: with the matching pack the deep tier scores higher",
   "value": "above_min 0.609 (0.532–0.679) in-pack n 32 vs 0.467 (0.417–0.519) not-in-pack n 78",
   "ci": "1 seed",
   "n": null,
   "seeds": "1",
   "set": "v0.2 dev (selection toward crypto/travel/danger)",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-12",
   "date_text": "2026-10-12 21:00",
   "source_public": null,
   "caveat": "P1 first reported a different floor scale (0.55/0.39); corrected",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-chron-late-19"
   ]
  },
  {
   "id": "F-286",
   "kind": "fact",
   "claim": "Same single-pack effect under Opus",
   "value": "+0.042; above_min +0.053",
   "ci": "(+0.024…+0.062); n 180",
   "n": "180",
   "seeds": null,
   "set": "the 180 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-13",
   "date_text": "2026-10-13 21:11",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:87",
   "caveat": "topic half +0.035 (+0.009…+0.062) clears zero only at L2",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "F-chron-late-27"
   ]
  },
  {
   "id": "F-287",
   "kind": "event",
   "claim": "Claim draft v4.2.4: Claude Opus stays the headline (lower judge), win-rate first",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "T-results-L76"
   ]
  },
  {
   "id": "F-288",
   "kind": "null-result",
   "claim": "Local open-weights judge as headline: Gemma 4 26B Q2 on the judge.v1 prompt",
   "value": "+0.11 lenient, κ 0.32-0.55, 41-43 s/call",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: not a drop-in ratio judge; ranking only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "N-method-L215"
   ]
  },
  {
   "id": "F-289",
   "kind": "null-result",
   "claim": "Raw ratio as headline: 2026-10-02 plan",
   "value": "floor gives ≈ 0.31 free",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "citation fix (cited l.13/63 are other rows). Note: the floor-adjusted scale was already a pre-registered companion metric in the 10-02 plan (analysis-plan.md:22-25); only its promotion to headline came on 10-06",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "floor-trap",
   "aliases": [
    "N-method-L219"
   ]
  },
  {
   "id": "F-290",
   "kind": "decision",
   "claim": "Seven axes (quality, truthfulness, safety, retrieval, latency, resources, stability), never averaged into one score (why: one average hides safety and latency regressions)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "source_public": null,
   "caveat": "outcome: quality ratio became the published number; other axes reported separately",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "D-method-02"
   ]
  },
  {
   "id": "F-291",
   "kind": "event",
   "claim": "exp-06: the app's real answer engine runs on the desktop (Mac mini), 54 questions in 2 min 34 s",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L38"
   ]
  },
  {
   "id": "F-292",
   "kind": "fact",
   "claim": "First 4B baseline on the shared engine (seed 42)",
   "value": "raw 0.521",
   "ci": "(0.467–0.574), n≈57, 1 seed",
   "n": "57,",
   "seeds": "1",
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-01",
   "source_public": null,
   "caveat": "floor-adj not given; CI contains 0.50",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-12"
   ]
  },
  {
   "id": "F-293",
   "kind": "event",
   "claim": "First frozen criterion before opening a blind set (embedding danger detector; sha frozen, \"no tuning after this point\")",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "date_text": "2026-10-01 21:17-21:30",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-process-L26"
   ]
  },
  {
   "id": "F-294",
   "kind": "event",
   "claim": "First frozen criterion in the queue (embedding danger detector, frozen before the blind set is opened)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "date_text": "2026-10-01 21:30",
   "source_public": null,
   "caveat": "Verification note: The 21:30 freeze is criterion #4 (QUEUE:21); line refs 24-29 are off by ~3. Process notes cite QUEUE:32-40, which is the later #4b freeze at 22:05.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L41"
   ]
  },
  {
   "id": "F-295",
   "kind": "event",
   "claim": "Embedding danger detector criterion frozen before its blind set exists (start of learned-detector line)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "date_text": "2026-10-01 21:30",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-system-L25"
   ]
  },
  {
   "id": "F-296",
   "kind": "event",
   "claim": "First 4B baseline on the shared engine: raw 0.521 (s42), \"first point above 0.50, CI contains it\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "date_text": "2026-10-01 23:40",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L43"
   ]
  },
  {
   "id": "F-297",
   "kind": "event",
   "claim": "Direction: \"eval + pipeline technique\" (better answers from the same model)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 evening",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L57"
   ]
  },
  {
   "id": "F-298",
   "kind": "fact",
   "claim": "At 10-02 the bounty had 4 claims, including the app's own earlier claim (1.5B model)",
   "value": "4 claims",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "n/a",
   "set_status": null,
   "judge": "n/a",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "claim names/handles not for publication",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-03"
   ]
  },
  {
   "id": "F-299",
   "kind": "fact",
   "claim": "Danger detector + no-source card (A1) cut danger memory answers on the 4B blind sets from 14 → 1 and 16 → 2, 0 sources lost",
   "value": "14→1, 16→2",
   "ci": "1 seed (s42)",
   "n": null,
   "seeds": "1",
   "set": "blind health sets (later dev)",
   "set_status": "dev",
   "judge": "rules + hand read",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "one real regression (CO \"leave the area\" lost)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-16"
   ]
  },
  {
   "id": "F-300",
   "kind": "fact",
   "claim": "Held-out precision budget: CI half-width 0.052 at 57 items, ~0.030 at 167",
   "value": "0.052 → 0.030",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "planning",
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-22"
   ]
  },
  {
   "id": "F-301",
   "kind": "fact",
   "claim": "One-token route gate on the 4B classified chat vs search at 95 % on a dev chat slice",
   "value": "95 % (false search 3/34, missed 0/23)",
   "ci": "57 scored",
   "n": null,
   "seeds": null,
   "set": "chat-slice v0 dev",
   "set_status": "dev",
   "judge": "deterministic labels",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "blind confirm later 55/57 (QUEUE:392)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-39"
   ]
  },
  {
   "id": "F-302",
   "kind": "event",
   "claim": "Danger detector + no-source card (A1) confirmed on both profiles; first safety win",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 02:15",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L45"
   ]
  },
  {
   "id": "F-303",
   "kind": "event",
   "claim": "Rubric-floor risk found: a blank system scores ~0.20–0.31 of the reference",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 06:10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L49"
   ]
  },
  {
   "id": "F-304",
   "kind": "decision",
   "claim": "Declare three health blind sets dev; commission fresh blind sets from writers with no file access — why: traces sat in readable paths and were used as dev — outcome: new sets health-fresh1/2/3, chat-v3, meta-fresh1",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "D-chron-early-16"
   ]
  },
  {
   "id": "F-305",
   "kind": "event",
   "claim": "Several health blind sets declared dev after traces sat in readable paths",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L64"
   ]
  },
  {
   "id": "F-306",
   "kind": "event",
   "claim": "Learned emergency gate generalizes on blind sets (90/90, 71/72); lexical does not",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L66"
   ]
  },
  {
   "id": "F-307",
   "kind": "event",
   "claim": "English only",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L69"
   ]
  },
  {
   "id": "F-308",
   "kind": "event",
   "claim": "Gold-passage oracle: 4B + gold passages 0.606 vs engine 0.507",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L73"
   ]
  },
  {
   "id": "F-309",
   "kind": "event",
   "claim": "Cut 1 / 1b built; cut 2 frozen (cff2f2bc…)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L74"
   ]
  },
  {
   "id": "F-310",
   "kind": "event",
   "claim": "Retrieval v2 killed: gold-in-context 7 → 18 yet quality −0.021",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L75"
   ]
  },
  {
   "id": "F-311",
   "kind": "fact",
   "claim": "Lexical health detector tuned on dev failed a fresh blind set: health route 31/72 (bar ≥ 54)",
   "value": "10 → 31/72",
   "ci": "1 seed",
   "n": null,
   "seeds": "1",
   "set": "health-fresh1 (blind then; dev now)",
   "set_status": "dev",
   "judge": "scripted",
   "date": "2026-10-03",
   "source_public": "boar-research:reports/SET-STATUS.md:16",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-40"
   ]
  },
  {
   "id": "F-312",
   "kind": "fact",
   "claim": "Health stack L2: unsafe answers 7 → 0 (4B) and 10 → 0 (1.5B) with no quality cost",
   "value": "ΔQ +0.009 (−0.009…+0.030)",
   "ci": "2 seeds",
   "n": null,
   "seeds": "2",
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus + safety rules v0.8.2",
   "date": "2026-10-03",
   "source_public": null,
   "caveat": "cards were unvetted drafts at the time",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-42"
   ]
  },
  {
   "id": "F-313",
   "kind": "fact",
   "claim": "The independent Opus reviewer changed 25 of 31 first-aid cards, rejected 0",
   "value": "6 approved / 25 changed",
   "ci": "31 cards",
   "n": null,
   "seeds": null,
   "set": "n/a",
   "set_status": null,
   "judge": "Opus reviewer",
   "date": "2026-10-03",
   "source_public": null,
   "caveat": "\"AI-reviewed, not clinically reviewed\"",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-43"
   ]
  },
  {
   "id": "F-314",
   "kind": "fact",
   "claim": "Oracle ladder: closed-book 0.487 · engine 0.507 · question-chosen passages 0.567 · gold passages 0.606 · 1500-token gold 0.626 · Haiku no-web 0.653",
   "value": "see value",
   "ci": "57 items",
   "n": null,
   "seeds": null,
   "set": "v0.1 EN dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-03",
   "source_public": null,
   "caveat": "gold passages chosen with knowledge of the reference → upper bound",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-46"
   ]
  },
  {
   "id": "F-315",
   "kind": "event",
   "claim": "Final cut 2 (cut 1b + exposure fix + citation fix)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "date_text": "2026-10-03/04",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-system-L32"
   ]
  },
  {
   "id": "F-316",
   "kind": "event",
   "claim": "Cut 3 adopted (cut 2 + self-contradiction guard + arithmetic tool + snakebite text)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L81"
   ]
  },
  {
   "id": "F-317",
   "kind": "fact",
   "claim": "4B loss anatomy: gap to Haiku +0.121 net; cut 2's gain over old main (+0.026) came entirely from non-model paths",
   "value": "+0.121",
   "ci": "57 items",
   "n": null,
   "seeds": null,
   "set": "v0.1 EN dev",
   "set_status": "dev",
   "judge": "existing verdicts",
   "date": "2026-10-04",
   "source_public": null,
   "caveat": "biggest buckets \"thin but correct\" and \"knowledge\" had no lever",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-54"
   ]
  },
  {
   "id": "F-318",
   "kind": "fact",
   "claim": "Cut 3 over cut 2 on v0.1",
   "value": "+0.032 CI90 [+0.010, +0.062]; 0.566 vs 0.534",
   "ci": "n 57",
   "n": "57",
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus judge.v1-med",
   "date": "2026-10-04",
   "source_public": null,
   "caveat": "last 4B-side adopted quality gain | rests on 4 items; did not transfer to held-out (F-chron-early-62)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-results-100",
    "F-chron-early-55"
   ]
  },
  {
   "id": "F-319",
   "kind": "fact",
   "claim": "Cut 2's gain over old main came from non-model paths",
   "value": "+0.028 cards + calculator; model-side ≈ 0",
   "ci": "n 57",
   "n": "57",
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "not stated at the cited line (neighbouring buckets are in Haiku-rubric terms)",
   "date": "2026-10-04",
   "source_public": null,
   "caveat": "citation fix: the +0.028 non-model-path finding is l.7",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-results-101"
   ]
  },
  {
   "id": "F-320",
   "kind": "event",
   "claim": "4B loss anatomy: cut 2 gains over old main come only from non-model paths (cards, calculator); \"thin\" and \"knowledge\" buckets have no 4B lever",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-results-L27"
   ]
  },
  {
   "id": "F-321",
   "kind": "event",
   "claim": "Cut 3 adopted (cut 2 + verdict guard v3 + arithmetic repair + snakebite text); L1 Δ +0.029 [+0.007, +0.058]",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-results-L29"
   ]
  },
  {
   "id": "F-322",
   "kind": "event",
   "claim": "Cut 3 confirmed: v0.1 EN 0.566 vs cut 2 0.534, Δ +0.032 CI90 [+0.010, +0.062]",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-results-L30"
   ]
  },
  {
   "id": "F-323",
   "kind": "event",
   "claim": "Computed-verdict line (CV/CVS/CVH/CVO) and safety-card trigger v1–v3 all closed at gates",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "source_public": "boar-research:reports/CLOSED.md:82",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-results-L31"
   ]
  },
  {
   "id": "F-324",
   "kind": "event",
   "claim": "Cut 3 adopted = cut 2 + verdict guard v3 + arithmetic tools + snakebite text",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-system-L34"
   ]
  },
  {
   "id": "F-325",
   "kind": "event",
   "claim": "Overnight generation: cut 2 v0.1 EN ≈ 0.533 over 4 seeds",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "date_text": "2026-10-04 01:03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L78"
   ]
  },
  {
   "id": "F-326",
   "kind": "event",
   "claim": "Phone latency root cause: our own KV-cache save-path bug (desktop shim hid it)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L85"
   ]
  },
  {
   "id": "F-327",
   "kind": "event",
   "claim": "First independent claim-text review: 21 issues",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L88"
   ]
  },
  {
   "id": "F-328",
   "kind": "fact",
   "claim": "Detector across turns (DETMAX) adopted into cut 6",
   "value": "follow-up mismatches 0/60 per prefix (control 7 total); false fires 1",
   "ci": "240 forms",
   "n": null,
   "seeds": null,
   "set": "dev probes",
   "set_status": "dev",
   "judge": "none",
   "date": "2026-10-05",
   "date_text": "2026-10-05/06",
   "source_public": "boar-research:arms/detfu/DETMAX.md:8-14",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-results-99"
   ]
  },
  {
   "id": "F-329",
   "kind": "event",
   "claim": "Claim runs on cut 3 approved and run: v0.2 held-out raw 0.513 (0.479–0.550); V-para 0.413 FAIL",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-results-L32"
   ]
  },
  {
   "id": "F-330",
   "kind": "event",
   "claim": "Claim runs on cut 3 (held-out v0.2 + V-para)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-system-L40"
   ]
  },
  {
   "id": "F-331",
   "kind": "decision",
   "claim": "Freeze the claim build at cut 6; later only routing/safety fixes that leave v0.1 answers unchanged — why: stable claim evidence — outcome: cut 7 later added the fact table (0 fires on held-out sets)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "D-chron-early-23"
   ]
  },
  {
   "id": "F-332",
   "kind": "decision",
   "claim": "Keep FACTS r6 as head despite FAIL by frozen rule (recall 70 % < 75 %) (why: r6 dominates r5 on every frozen measure; precision 96.6 %, FP 0/100 adversarial)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "date_text": "2026-10-06/07",
   "source_public": "boar-research:arms/facts/FACTS-r6-SET4.md:25-28",
   "caveat": "outcome: Shipped in cut 7; stated as \"FAIL stands\"",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "D-results-03"
   ]
  },
  {
   "id": "F-333",
   "kind": "decision",
   "claim": "Adopt DETMAX → cut 6 (why: chit-chat before an emergency lost 7/240 routes; DETMAX 0/240)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "citation fix (l.46 is blank)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "D-system-08"
   ]
  },
  {
   "id": "F-334",
   "kind": "event",
   "claim": "Cut 5 (speed-only), cut 6 (detector fix)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L90"
   ]
  },
  {
   "id": "F-335",
   "kind": "event",
   "claim": "Claim build frozen at cut 6; second held-out exposure",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L92"
   ]
  },
  {
   "id": "F-336",
   "kind": "event",
   "claim": "Claim runs (cut 6): v0.2 raw 0.512 (two seeds), V-para 0.413",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L94"
   ]
  },
  {
   "id": "F-337",
   "kind": "event",
   "claim": "FACTS r4–r6 + blind recall sets: precision solved, recall 60–70 %; line closed, r6 kept",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "date_text": "2026-10-06/07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L97"
   ]
  },
  {
   "id": "F-338",
   "kind": "fact",
   "claim": "On cut 6 v0.2 the 4B's correctness was 3.4/5 but completeness 2.07 and usefulness 2.2; median 53 words vs reference ≤ 250",
   "value": "3.4 / 2.07 / 2.2",
   "ci": "110 items",
   "n": null,
   "seeds": null,
   "set": "v0.2",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": "boar-research:reports/BAR-TABLE.md:90",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-74"
   ]
  },
  {
   "id": "F-339",
   "kind": "fact",
   "claim": "FACTS (Wikidata fact table, no model) on its 8 fired v0.1 items: +0.94 rubric per item, projected v0.1 Δ ≈ +0.027",
   "value": "+0.94/item",
   "ci": "16 calls, 8 items",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": "boar-research:arms/facts/FACTS-JUDGE.md",
   "caveat": "dev-motivated; 0 fires on v0.2 and V-para",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-78"
   ]
  },
  {
   "id": "F-340",
   "kind": "fact",
   "claim": "Judge triage calibration: Sonnet-low picked the same winner as Opus on 54/57, ratio −0.04 systematic, 3.1 s/call",
   "value": "54/57",
   "ci": "57 items",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Sonnet vs Opus",
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "replaced by Haiku next day",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-80"
   ]
  },
  {
   "id": "F-341",
   "kind": "fact",
   "claim": "An app that declines every question scores raw ≈ 0.31, i.e. ≈ 61 % of the 0.50 bar for free.",
   "value": "0.306 / 61 %",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.1",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": "boar-bench:TRACE-FORMAT.md:67",
   "caveat": "citation fix: l.13 is the cut 6 v0.2 row",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-method-14"
   ]
  },
  {
   "id": "F-342",
   "kind": "fact",
   "claim": "Cut 6 vs cut 3 on v0.2 is no change",
   "value": "Δ −0.0009",
   "ci": "−0.0095…+0.0077; n 110",
   "n": "110",
   "seeds": null,
   "set": "v0.2 held-out at time",
   "set_status": "held-out at run time",
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "citation fix (l.48 is a separator). 323 of 334 verdicts were copied from cut 3 for byte-identical answers, so the narrow Δ reflects reuse, not an independent re-judging",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-results-07"
   ]
  },
  {
   "id": "F-343",
   "kind": "fact",
   "claim": "FACTS table (offline fact lookup) blind recall / precision",
   "value": "recall 28/40 = 70 %; precision 96.6 %; false positives 0/20 (0/100 across adversarial sets)",
   "ci": "n 40",
   "n": "40",
   "seeds": null,
   "set": "blind set 4 (burnt)",
   "set_status": "burnt → dev",
   "judge": "none (deterministic)",
   "date": "2026-10-06",
   "source_public": "boar-research:arms/facts/FACTS-r6-SET4.md:4-9",
   "caveat": "v0.1 effect projected +0.027 (8 fires)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-results-97"
   ]
  },
  {
   "id": "F-344",
   "kind": "fact",
   "claim": "FACTS table size",
   "value": "4,233 entities, 761 KB",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": "boar-research:arms/facts/FACTS-r6.md:5",
   "caveat": "Wikidata CC0",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-results-98"
   ]
  },
  {
   "id": "F-345",
   "kind": "fact",
   "claim": "DETMAX fixes chit-chat prefixes before an emergency.",
   "value": "lost routes 7/240 → 0/240; split-turn forms lose 0/60; +1 route in 192 dev chats (a real flood emergency)",
   "ci": "n 240 / 60 / 192",
   "n": "240",
   "seeds": null,
   "set": "dev probes",
   "set_status": "dev",
   "judge": "none",
   "date": "2026-10-06",
   "source_public": "boar-research:arms/detfu/DETMAX.md",
   "caveat": "citation fix (l.46 blank)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-system-06"
   ]
  },
  {
   "id": "F-346",
   "kind": "fact",
   "claim": "FACTS asset size.",
   "value": "4,233 entities, 760,970 B",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": "boar-research:arms/facts/FACTS-r6.md:5",
   "caveat": "citation fix (l.6 is the sha line)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-system-19"
   ]
  },
  {
   "id": "F-347",
   "kind": "fact",
   "claim": "FACTS r6 on a blind recall set.",
   "value": "recall 28/40 = 70.0 % (FAIL < 75 %), precision 28/29 = 96.6 %; adversarial FP 0/100",
   "ci": "n 40",
   "n": "40",
   "seeds": null,
   "set": "blind set 4 (burnt after)",
   "set_status": "burnt → dev",
   "judge": "none",
   "date": "2026-10-06",
   "source_public": "boar-research:arms/facts/FACTS-r6-SET4.md; boar-research:arms/facts/FACTS-r6-SET4.md:1-8,27-28",
   "caveat": "shipped anyway as cut 7 head; recall 60–70 % on fresh phrasing",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-system-20"
   ]
  },
  {
   "id": "F-348",
   "kind": "fact",
   "claim": "FACTS projected effect on v0.1.",
   "value": "+0.027 (8 fires, +0.94 rubric/item)",
   "ci": "8 items, judged once",
   "n": null,
   "seeds": null,
   "set": "v0.1 (dev)",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": "boar-research:arms/facts/FACTS-JUDGE.md; boar-research:arms/facts/FACTS-JUDGE.md:14",
   "caveat": "dev-motivated item selection",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-system-21"
   ]
  },
  {
   "id": "F-349",
   "kind": "event",
   "claim": "Cut 5 = cut 4 + no-source KV prefix + route-gate skip (CUT5-LITE adopted; CUT5-SPEED rejected on desktop gate)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-system-L42"
   ]
  },
  {
   "id": "F-350",
   "kind": "event",
   "claim": "Cut 6 = cut 5 + DETMAX (detector scores window and current turn)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-system-L43"
   ]
  },
  {
   "id": "F-351",
   "kind": "event",
   "claim": "Pipeline audit: \"rigorous process, looping direction\"; honest metric adopted",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "date_text": "2026-10-06 14:06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L93"
   ]
  },
  {
   "id": "F-352",
   "kind": "event",
   "claim": "FACTS r6 on blind set 4: recall 70 % < 75 % → FAIL; kept as head by decision",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "date_text": "2026-10-06 23:54",
   "source_public": "boar-research:arms/facts/FACTS-r6-SET4.md:1-9",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-results-L37"
   ]
  },
  {
   "id": "F-353",
   "kind": "decision",
   "claim": "Adopt cut 7 = cut 6 + FACTS r6 on gates alone (why: FACTS fires 0/110 v0.2, 0/58 V-para → metrics unchanged; product gain (instant table answers))",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": "boar-research:arms/facts/FACTS-r6-SET4.md:1,24-28",
   "caveat": "outcome: cut 7 = frozen claim build; FACTS recall line closed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "D-system-10"
   ]
  },
  {
   "id": "F-354",
   "kind": "event",
   "claim": "Fine-tune data line closed (teacher 16/20 → +rejection 11/20 → human 6/20 unsupported)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L100"
   ]
  },
  {
   "id": "F-355",
   "kind": "fact",
   "claim": "FACTS final state on blind recall sets: precision 90–97 %, 0/100 adversarial false positives, recall 60–70 %",
   "value": "recall 28/40 (r6, set 4)",
   "ci": "40 + 20 items per set",
   "n": null,
   "seeds": null,
   "set": "blind recall sets (spent)",
   "set_status": null,
   "judge": "deterministic",
   "date": "2026-10-07",
   "source_public": "boar-research:arms/facts/FACTS-r6-SET4.md",
   "caveat": "failed its frozen 75 % recall bar; shipped anyway as cut 7 head",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-79"
   ]
  },
  {
   "id": "F-356",
   "kind": "fact",
   "claim": "Cut 7 fact table fires 0/110 on v0.2 and 0/58 on V-para → held-out numbers are cut 6's",
   "value": "0 fires",
   "ci": "deterministic",
   "n": null,
   "seeds": null,
   "set": "v0.2 / V-para",
   "set_status": null,
   "judge": "n/a",
   "date": "2026-10-07",
   "source_public": "boar-research:reports/BAR-TABLE.md:29",
   "caveat": "only measured effect: v0.1 +0.027 projected",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-83"
   ]
  },
  {
   "id": "F-357",
   "kind": "fact",
   "claim": "Judge sharding N = 4: ×4.5 throughput, 0 rate limits",
   "value": "9.9 → 44.4 calls/min",
   "ci": "114 calls/level",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": "Haiku",
   "date": "2026-10-07",
   "source_public": "boar-research:reports/RULE-CHANGES.md:10",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-87"
   ]
  },
  {
   "id": "F-358",
   "kind": "fact",
   "claim": "Cut 7 build identity.",
   "value": "commit 6d0b93b1, tree 7c96e59e…, patch vs cut 6 sha 0eee6a0b…; 1,346/1,346 unit tests",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": "local commit, never pushed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-system-17"
   ]
  },
  {
   "id": "F-359",
   "kind": "fact",
   "claim": "FACTS (Wikidata table) fires zero times on the held-out sets, so cut 7 = cut 6 there.",
   "value": "0/110 v0.2, 0/58 V-para; 8/72 on v0.1",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "v0.2 (burnt→dev), V-para",
   "set_status": "burnt → dev",
   "judge": "none",
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-system-18"
   ]
  },
  {
   "id": "F-360",
   "kind": "event",
   "claim": "Cut 7 = cut 6 + FACTS r6 adopted on gates alone (0 fires on held-out)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-system-L47"
   ]
  },
  {
   "id": "F-361",
   "kind": "event",
   "claim": "Cut 7 frozen = cut 6 + FACTS r6 (adopted on gates; fires 0/110 v0.2, 0/58 V-para)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "date_text": "2026-10-07 09:32",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-results-L40"
   ]
  },
  {
   "id": "F-362",
   "kind": "event",
   "claim": "Autonomous research mandate",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "date_text": "2026-10-07 14:26",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L101"
   ]
  },
  {
   "id": "F-363",
   "kind": "fact",
   "claim": "Thinking budget 3072 added nothing: the model stops by itself at ~1.3k tokens",
   "value": "Δ +0.010 (−0.014…+0.041)",
   "ci": "1 seed",
   "n": null,
   "seeds": "1",
   "set": "V-para dev-style",
   "set_status": "dev-style",
   "judge": "Haiku L1",
   "date": "2026-10-08",
   "source_public": "boar-research:reports/CLOSED.md:41",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-early-93"
   ]
  },
  {
   "id": "F-364",
   "kind": "event",
   "claim": "Thinking budget curve: 512 = 1024 at 0.55× time; 256 loses; knee at 512",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L109"
   ]
  },
  {
   "id": "F-365",
   "kind": "fact",
   "claim": "References are ≈ 280 words (p10/p50/p90 = 251/280/318); the 4B app's answers median ≈ 51-53 words.",
   "value": "280 vs ~52 words",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para / v0.2",
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:95",
   "caveat": "citation fix: 53 words (cut 6 v0.2) at PIPELINE-AUDIT l.17, 51 words (v0.1) at RND-REVIEW l.50; 251/280/318 at BAR-TABLE l.95",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-method-31"
   ]
  },
  {
   "id": "F-366",
   "kind": "event",
   "claim": "R&D process review: strong instrument, wrong target; deliverables unbuilt",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 13:30",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L113"
   ]
  },
  {
   "id": "F-367",
   "kind": "event",
   "claim": "Author: \"stop everything\"; then \"research on the Mac mini only until the bar is met\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 21:50",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-early-L116"
   ]
  },
  {
   "id": "F-368",
   "kind": "decision",
   "claim": "Licence policy: PD, CC0, CC BY, CC BY-SA only; never NC, NHS, A.D.A.M.; SE from archive.org 2024-04-02 snapshot (why: redistribution clarity)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": null,
   "caveat": "citation fix: D-K5 is l.12 (l.13 is D-K6)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "D-system-14"
   ]
  },
  {
   "id": "F-369",
   "kind": "fact",
   "claim": "NPU prefill does not help this model.",
   "value": "≈ 0 s net; UD-Q2_K_XL experts not taken by the NPU",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": "boar-research:reports/DEEP-RESEARCH-2026-10-12.md:25",
   "caveat": "third-party engine data",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-system-51"
   ]
  },
  {
   "id": "F-370",
   "kind": "event",
   "claim": "Knowledge Pack catalogue and authoring workflow (D-K1..D-K7)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 23:4x",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-system-L52"
   ]
  },
  {
   "id": "F-371",
   "kind": "fact",
   "claim": "Resume safety: a cached judgment row is reused only if traces, questions and references sha256, judge version and model all match; else exit 7.",
   "value": "exit 7",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": "boar-bench:harness/judge.mjs:11-14; boar-bench:TRACE-FORMAT.md:56",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-method-77"
   ]
  },
  {
   "id": "F-372",
   "kind": "event",
   "claim": "Wave 2 published (factbook, simplewiki)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 02:30",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-system-L54"
   ]
  },
  {
   "id": "F-373",
   "kind": "event",
   "claim": "PR-1 (cut 7 rebased on upstream + chat wiring) engineering complete, review \"mergeable as is\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 22:31–22:54",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-late-L73"
   ]
  },
  {
   "id": "F-374",
   "kind": "decision",
   "claim": "Health answers never come from the model on the default tier; doses never from a model (HEALTHGAP); a fired detector can only be overridden by a deterministic calculator (NUMGUARD r3) (why: safety; model answered doses from memory on probes)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15/16",
   "source_public": null,
   "caveat": "citation fix: 'never reach the 4B' is l.21. AI-reviewed, not clinically reviewed",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "D-system-18"
   ]
  },
  {
   "id": "F-375",
   "kind": "fact",
   "claim": "Health questions on the default tier never reach the 4B: they get a fixed card or a verbatim source excerpt (0 model tokens).",
   "value": "modelId grounding-guard / extractive",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": "citation fix; design statement; AI-reviewed, not clinically reviewed",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-system-07"
   ]
  },
  {
   "id": "F-376",
   "kind": "fact",
   "claim": "First-aid card set size and review status.",
   "value": "32 card ids (31 in JSON × EN/PT/ES + snakebite in code), 75,955 B JSON, all \"reviewed-by-opus-reviewer\"",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": "citation fix (card data is l.61); AI-reviewed, not clinically reviewed",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-system-08"
   ]
  },
  {
   "id": "F-377",
   "kind": "fact",
   "claim": "Health excerpts are verbatim, at most 700 characters, and pass a risky-instruction filter (e.g. tourniquet, suck venom, butter, head back for nosebleed).",
   "value": "700 chars; ~13 risky ids",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": "citation fix (≤ 700 chars l.16, risky-instruction list l.105); the '~13 ids' count was not re-counted",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-system-09"
   ]
  },
  {
   "id": "F-378",
   "kind": "fact",
   "claim": "HEALTHGAP: 15 of 64 health-shaped probes reached the model because the detector scored them 0.00–0.16 (dose, poison, CO reading, survival time).",
   "value": "15/64",
   "ci": "n 64 probes",
   "n": "64",
   "seeds": null,
   "set": "agent-written dev probes",
   "set_status": "dev",
   "judge": "none",
   "date": "2026-10-15",
   "source_public": null,
   "caveat": "citation fix (l.156-158 are open questions)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-system-13"
   ]
  },
  {
   "id": "F-380",
   "kind": "fact",
   "claim": "Per-mechanism share of that gap",
   "value": "first-aid cards 4 items +0.338 (≈ 27 %); facts cards 8 items +0.166 (≈ 29 %); 4B + grounding 42 items +0.038 (≈ 30 %); calculator 2 items +0.607 (≈ 22 %)",
   "ci": "n 57",
   "n": "57",
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Haiku L1",
   "date": "2026-10-15",
   "date_text": "2026-10-15 15:17",
   "source_public": null,
   "caveat": "corrected breakdown; posted as relative shares only | Verification note: Health/grounding/calculator shares are at QUEUE:1156 (first breakdown); the card split (27 % / 29 %) is the 15:17 correction at :1162. Shares sum > 100 % because one extractive item is −0.333. Internal-only per the author.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-late-55"
   ]
  },
  {
   "id": "F-381",
   "kind": "event",
   "claim": "MATHROUTE: math-through-the-model router does not recover the calculator (20/26 vs 23/26)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 15:24",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-late-L80"
   ]
  },
  {
   "id": "F-382",
   "kind": "fact",
   "claim": "Math through the model vs calculator (MATHROUTE)",
   "value": "cut 7 23/26 vs model-only router 20/26",
   "ci": "n 26 (2 v0.1 + 24 agent-written)",
   "n": "26",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "deterministic scorer",
   "date": "2026-10-15",
   "date_text": "2026-10-15 15:24",
   "source_public": null,
   "caveat": "the detector refused 3–5/26 plain numeric questions",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-late-62"
   ]
  },
  {
   "id": "F-383",
   "kind": "event",
   "claim": "Native iOS port started (SwiftUI + Swift port of the Kotlin engine)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 19:5x",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "T-system-L62"
   ]
  },
  {
   "id": "F-384",
   "kind": "event",
   "claim": "NUMGUARD r1→r3: lexicon guard rejected as unsafe by review; calculator-only r3 PASS",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 20:36–21:56",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-late-L84"
   ]
  },
  {
   "id": "F-385",
   "kind": "event",
   "claim": "Independent reviewer REJECTs a numeric guard as unsafe; scope cut to calculator-only",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 21:34",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-process-L67"
   ]
  },
  {
   "id": "F-386",
   "kind": "fact",
   "claim": "Calculator-only numeric guard (NUMGUARD r3)",
   "value": "7/7 checks pass; 0 health probes leak",
   "ci": "26-item slice + 32/6/18/2 probes",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "none",
   "date": "2026-10-15",
   "date_text": "2026-10-15 21:56",
   "source_public": null,
   "caveat": "a regression gate, not evidence of a quality effect (CRITIQUE row 11)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-late-63"
   ]
  },
  {
   "id": "F-387",
   "kind": "event",
   "claim": "HEALTHGAP: 15 health probes reach the model unguarded; predicate v1.1/v1.2, implementation PASS",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 22:51–2026-10-16 00:58",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "E-chron-late-L87"
   ]
  },
  {
   "id": "F-388",
   "kind": "fact",
   "claim": "Health questions that reached the model unguarded (HEALTHGAP)",
   "value": "15/64 probes; predicate covers 13/15; fires 1/280 public",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "agent-written probes",
   "set_status": null,
   "judge": "none",
   "date": "2026-10-15",
   "date_text": "2026-10-15 22:51",
   "source_public": null,
   "caveat": "probes written next to the rule (in-sample)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-late-64"
   ]
  },
  {
   "id": "F-389",
   "kind": "decision",
   "claim": "deps.warn (unsafe-answer warning) stays off by default — why: out-of-sample recall 3/8 (lexicon) and 12/18 with FP 8/92 (detector) — outcome: not handed to the PR",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "D-chron-late-36"
   ]
  },
  {
   "id": "F-390",
   "kind": "fact",
   "claim": "Lite tier safety.",
   "value": "health routing identical to 4B 124/124; model answers alone 3/64 health probes; unsafe 1 vs 1",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "agent-written probes + v0.1",
   "set_status": null,
   "judge": "jev + rules",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "claim kit only, one seed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-system-80"
   ]
  },
  {
   "id": "F-391",
   "kind": "fact",
   "claim": "Deterministic layers answer most health probes on the lite path.",
   "value": "61/64 probes answered by deterministic layers",
   "ci": "n 64",
   "n": "64",
   "seeds": null,
   "set": "agent-written probes",
   "set_status": null,
   "judge": "none",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-system-90"
   ]
  },
  {
   "id": "F-392",
   "kind": "fact",
   "claim": "Unsafe-answer warning out of sample",
   "value": "lexicon recall 3/8 (in-sample 15/15); detector-based 12/18 with FP 8/92",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.2 dev, script-only",
   "set_status": "dev",
   "judge": "none",
   "date": "2026-10-16",
   "date_text": "2026-10-16 13:14",
   "source_public": null,
   "caveat": "warning stays off by default",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-late-71"
   ]
  },
  {
   "id": "F-393",
   "kind": "fact",
   "claim": "Emergency-detector false fires on non-health v0.2 items",
   "value": "7 false fires; cost at most +0.019 raw",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.2 dev",
   "set_status": "dev",
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 13:20",
   "source_public": null,
   "caveat": "below MDE: closed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "F-chron-late-72"
   ]
  },
  {
   "id": "F-395",
   "kind": "null-result",
   "claim": "Wikivoyage lead gate (#5): drop off-topic travel guides → created an unsafe memory answer (frostbite) (gate FAIL)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L282"
   ]
  },
  {
   "id": "F-396",
   "kind": "null-result",
   "claim": "Title pseudo-relevance feedback (#10b): symptom → condition article → still a memory answer; side losses (FAIL)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L283"
   ]
  },
  {
   "id": "F-397",
   "kind": "null-result",
   "claim": "Longer-answer persona (#12): \"4–6 sentences\" → Δ −0.044; CW 10 → 20 (quality and safety worse)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L284"
   ]
  },
  {
   "id": "F-398",
   "kind": "null-result",
   "claim": "Self-harm crisis card v0/v1 (#16): detector + safe-messaging card → 25/30 blind (bar 85 %) (below bar; parked by the author as a known gap)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L285"
   ]
  },
  {
   "id": "F-399",
   "kind": "null-result",
   "claim": "NLI passage checker (P3-NLI): mDeBERTa contradiction check → 0/7 hits (wrong claims are memory, not passage contradictions)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L290"
   ]
  },
  {
   "id": "F-400",
   "kind": "null-result",
   "claim": "Doc2Query--: expand index with synthetic queries → Δ keys 0 (post-BM25 gate is the bottleneck)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L291"
   ]
  },
  {
   "id": "F-401",
   "kind": "null-result",
   "claim": "Native tool calling for routing: model decides to search → missed 20–22/23 searches (answers from memory)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L293"
   ]
  },
  {
   "id": "F-402",
   "kind": "null-result",
   "claim": "Reply-language arms (PT/ES): language line, precedence → hand-read failures (killed; then all PT/ES parked (English only))",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L294"
   ]
  },
  {
   "id": "F-403",
   "kind": "null-result",
   "claim": "Lexical health detector: tuned regex → 31/72 on blind (bar 54) (does not generalize)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L295"
   ]
  },
  {
   "id": "F-404",
   "kind": "null-result",
   "claim": "Premise guard: correct false premises → withheld 4 correct corrections/seed (KILL)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L299"
   ]
  },
  {
   "id": "F-405",
   "kind": "null-result",
   "claim": "Answer style S2/S3: answer-first, check-then-conclude → −0.031 / −0.006; CW +8 (style lever closed)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L300"
   ]
  },
  {
   "id": "F-406",
   "kind": "null-result",
   "claim": "Water/snakebite card trigger (SC v1–v3): new safety cards → false fires 5/50; recall 72 % (3 rounds failed; parked)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L303"
   ]
  },
  {
   "id": "F-407",
   "kind": "null-result",
   "claim": "Self-contradiction guard v4 (VG v4): more rules → 1 WRONG fire on fresh probes; catches 1/17 (line closed at v3)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L304"
   ]
  },
  {
   "id": "F-408",
   "kind": "null-result",
   "claim": "Travel section selection (TRAVELSEL): Wikivoyage practical sections → 1/3 targets (FAIL)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L309"
   ]
  },
  {
   "id": "F-409",
   "kind": "null-result",
   "claim": "Calculator coverage: more calculators → ceiling +0.000 on v0.1 (SKIP)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L310"
   ]
  },
  {
   "id": "F-410",
   "kind": "null-result",
   "claim": "Health follow-up routing (HFU-FIX/2): route follow-ups to health → 84 % (bar 90 %) (below bar; logged as upstream bug)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L315"
   ]
  },
  {
   "id": "F-411",
   "kind": "null-result",
   "claim": "Crypto truthfulness (CRYPTOTRUTH): honest not-found + EIP card → Δ +0.009; 1/3 fictional probes FAIL (research-only, parked)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L316"
   ]
  },
  {
   "id": "F-412",
   "kind": "null-result",
   "claim": "Wikidata facts recall (FACTS r4–r6): broader triggers, more entities → recall 60–70 % on fresh blind sets (missed 75 % bar; line closed, r6 shipped)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:24",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L319"
   ]
  },
  {
   "id": "F-413",
   "kind": "null-result",
   "claim": "Relaxed lowercase trigger (LCRELAX): allow lowercase names → +10 pts recall, unmeasured FP risk (not adopted)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L320"
   ]
  },
  {
   "id": "F-414",
   "kind": "null-result",
   "claim": "Crypto facts table (CRYPTOFACTS): structured crypto facts → 2/11 items lookup-able (coverage too low)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-early-L323"
   ]
  },
  {
   "id": "F-415",
   "kind": "null-result",
   "claim": "MP4B: the engine's own multi-pass decompose→research→synthesize on the 4B → Haiku −0.007 (−0.036…+0.023), ×4 time (futility rule; last untested 4B mechanism)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-late-L258"
   ]
  },
  {
   "id": "F-416",
   "kind": "null-result",
   "claim": "A3 math router: send math to the model without retrieval or calculator → 20/26 vs 23/26 (does not recover the calculator)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-late-L288"
   ]
  },
  {
   "id": "F-417",
   "kind": "null-result",
   "claim": "NUMGUARD lexicon guard: regex guard before the health refusal → 16/32 numeric health probes leak (unsafe; replaced by calculator-only r3)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-chron-late-L289"
   ]
  },
  {
   "id": "F-418",
   "kind": "null-result",
   "claim": "Units/time-zone/percent calculator (example of a 0-call ceiling): sized before building",
   "value": "Δ ceiling +0.000 on v0.1 (0/57 targets); 3/1,470 dev questions",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: ceiling below MDE; skip",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-method-L222"
   ]
  },
  {
   "id": "F-419",
   "kind": "null-result",
   "claim": "Lexicon numeric/health guard: NUMGUARD rounds r1-FINAL → reviewer REJECT (\"cannot be made safe\") (Scope cut to calculator-only (r3))",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-process-L209"
   ]
  },
  {
   "id": "F-420",
   "kind": "null-result",
   "claim": "R2 Gemma 4 E4B as reader: full engine, 2 seeds",
   "value": "Δ −0.0385 (CI90 −0.073…−0.006); 1.49× slower",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:17",
   "caveat": "why closed: closed (stop rule)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-results-L229"
   ]
  },
  {
   "id": "F-421",
   "kind": "null-result",
   "claim": "R9 FACTS instant tier: Wikidata table r1–r6",
   "value": "recall 70 % < 75 % bar",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:24",
   "caveat": "why closed: FAIL by frozen rule; shipped by decision",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-results-L236"
   ]
  },
  {
   "id": "F-422",
   "kind": "null-result",
   "claim": "R41 Numeric/date check vs FACTS table: jev gate",
   "value": "table covers 0/16, 0/17",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:65",
   "caveat": "why closed: step-0 FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-results-L268"
   ]
  },
  {
   "id": "F-423",
   "kind": "null-result",
   "claim": "R58–R60 Safety-card trigger v1–v3: 3 fresh sets",
   "value": "water recall 72 %/84 %/72 %; false fires 0/5/2",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:87",
   "caveat": "why closed: each FAIL; water card parked",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-results-L282"
   ]
  },
  {
   "id": "F-424",
   "kind": "null-result",
   "claim": "R80 Router instead of calculator for math: 26 items",
   "value": "20/26 vs 23/26",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: closed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-results-L301"
   ]
  },
  {
   "id": "F-425",
   "kind": "null-result",
   "claim": "R81 Lexicon numeric guard before health refusal: gates + review",
   "value": "review: 8 of 20 new probes went to model closed-book",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: closed as unsafe (calculator-only r3 survives)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-results-L302"
   ]
  },
  {
   "id": "F-426",
   "kind": "null-result",
   "claim": "R86 Contextual chunk headers: replay",
   "value": "top-8 69 → 75 (bar 90)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: step-0 FAIL; lever already in engine",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-results-L307"
   ]
  },
  {
   "id": "F-427",
   "kind": "null-result",
   "claim": "Deep context 1200 / long persona: more context, \"detailed\" persona",
   "value": "−0.006; −0.066 / −0.013",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md#R23-R25",
   "caveat": "why closed: no gain",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-system-L211"
   ]
  },
  {
   "id": "F-428",
   "kind": "null-result",
   "claim": "Cross-encoder reranker: bge-reranker on deep path and X3",
   "value": "gold into top-6 on 2 items; ~3.7 s",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md",
   "caveat": "why closed: gate fail",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-system-L217"
   ]
  },
  {
   "id": "F-429",
   "kind": "null-result",
   "claim": "Water safety card trigger: 3 rounds on blind sets",
   "value": "recall 72–84 %, false fires up to 5/50",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md#R58-R60",
   "caveat": "why closed: gate fails; parked",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-system-L218"
   ]
  },
  {
   "id": "F-430",
   "kind": "null-result",
   "claim": "Health follow-up lexical fix: HFU-FIX, rescue",
   "value": "58.1 % / 84.2 % (bar 90 %)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md#R61-R62",
   "caveat": "why closed: not adopted; DETMAX solved the emergency case",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-system-L219"
   ]
  },
  {
   "id": "F-431",
   "kind": "null-result",
   "claim": "Country-facts lookup (CF): Wikidata 197 states",
   "value": "data agreement 84 % (bar 98 %), recall 80/100",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md#R50",
   "caveat": "why closed: gate fail; FACTS replaced it",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "builds",
   "aliases": [
    "N-system-L221"
   ]
  },
  {
   "id": "F-432",
   "kind": "event",
   "claim": "First frozen criteria on fresh sets (hm-fresh2/3) before any reading",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "T-results-L23"
   ]
  },
  {
   "id": "F-433",
   "kind": "fact",
   "claim": "Asking the 4B for longer answers (persona P2) lowered quality: correct items 39 % → 14 %, confident-wrong 10 → 20",
   "value": "Δ −0.044",
   "ci": "(−0.076…−0.014), MDE 0.029",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "judge length slope negative (−0.34 per log-word)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-chron-early-18"
   ]
  },
  {
   "id": "F-434",
   "kind": "fact",
   "claim": "Exact key-in-context metric showed the earlier article-mode count overstated ~2×",
   "value": "13/52 vs 24/52",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "iPhone profile v0.1",
   "set_status": null,
   "judge": "deterministic",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "earlier numbers = upper bounds",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-chron-early-19"
   ]
  },
  {
   "id": "F-435",
   "kind": "fact",
   "claim": "Grounded answer planning hurt the 4B, most where the kit had the facts",
   "value": "Δ −0.053",
   "ci": "(−0.085…−0.021), 2 seeds",
   "n": null,
   "seeds": "2",
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "second prompt-shaping failure",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-chron-early-33"
   ]
  },
  {
   "id": "F-436",
   "kind": "fact",
   "claim": "None of 7 wrong 4B sentences was contradicted by a passage: they were memory claims next to off-topic passages",
   "value": "0/7",
   "ci": "7 + 20 items",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "NLI model",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "closed passage checkers",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-chron-early-34"
   ]
  },
  {
   "id": "F-437",
   "kind": "fact",
   "claim": "Keep-the-key retrieval fix passed non-inferiority (+2 keys, 0 lost)",
   "value": "Δ −0.003",
   "ci": "(−0.019…+0.015)",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "first pipeline-technique \"win\", quality-neutral",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-chron-early-35"
   ]
  },
  {
   "id": "F-438",
   "kind": "fact",
   "claim": "4B vs 1.5B paired gap on v0.1 (pre-registered secondary): raw +0.131 (+0.095…+0.171), floor-adj +0.165.",
   "value": "+0.131 / +0.165",
   "ci": "2-3 seeds, n 57",
   "n": "57",
   "seeds": "3",
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": "boar-bench:bench/v0.3/plan/analysis-plan.md:69-74",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-method-92"
   ]
  },
  {
   "id": "F-439",
   "kind": "event",
   "claim": "Two-seed baseline: 4B raw 0.512 (0.461–0.567), 1.5B 0.379",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 01:00",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "E-chron-early-L44"
   ]
  },
  {
   "id": "F-440",
   "kind": "event",
   "claim": "Longer-answer persona makes the 4B hallucinate (Δ −0.044, confident-wrong 10 → 20)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 04:30",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "E-chron-early-L46"
   ]
  },
  {
   "id": "F-441",
   "kind": "fact",
   "claim": "A 4B one-token emergency gate generalized on two blind sets; lexical did not",
   "value": "90/90, HN 6/60; 71/72, HN 2/48 (lexical 23/90, 32/72)",
   "ci": "detection only",
   "n": null,
   "seeds": null,
   "set": "health-fresh2/1 (blind then; dev now)",
   "set_status": "dev",
   "judge": "scripted",
   "date": "2026-10-03",
   "source_public": "boar-research:reports/SET-STATUS.md:16-17",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-chron-early-41"
   ]
  },
  {
   "id": "F-442",
   "kind": "fact",
   "claim": "4B-tier model screen: Qwen3.5-9B closed-book = Qwen3-4B; Ministral 8B worse",
   "value": "0.488 vs 0.487; 0.457",
   "ci": "57 items, 1 seed",
   "n": null,
   "seeds": "1",
   "set": "v0.1 EN dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-03",
   "source_public": "boar-research:reports/CLOSED.md:16",
   "caveat": "\"not detected at MDE 0.055\"",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-chron-early-45"
   ]
  },
  {
   "id": "F-443",
   "kind": "fact",
   "claim": "Retrieval v2 put the gold sentence in context 7 → 18/56 times but lowered quality",
   "value": "Δ −0.021",
   "ci": "CI90 −0.054…+0.009, 2 seeds",
   "n": null,
   "seeds": "2",
   "set": "v0.1 EN dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-03",
   "source_public": null,
   "caveat": "retrieval 3.6× slower; 3 items carry the loss",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-chron-early-47"
   ]
  },
  {
   "id": "F-444",
   "kind": "fact",
   "claim": "Answer-style prompts all lost: answer-first −0.031, check-then-conclude −0.006, each raising confident-wrong",
   "value": "S2 0.468 / S3 0.493 vs 0.499",
   "ci": "1 seed",
   "n": null,
   "seeds": "1",
   "set": "v0.1 EN dev",
   "set_status": "dev",
   "judge": "Opus-med",
   "date": "2026-10-03",
   "source_public": null,
   "caveat": "one seed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-chron-early-52"
   ]
  },
  {
   "id": "F-445",
   "kind": "event",
   "claim": "Rule: no human review; an independent Opus reviewer decides wording; \"AI-reviewed, not clinically reviewed\" caveat",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": "boar-research:reports/RULE-CHANGES.md:7",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "T-results-L26"
   ]
  },
  {
   "id": "F-446",
   "kind": "event",
   "claim": "4B loss anatomy: gap to Haiku +0.121 net; biggest buckets have no live lever",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "E-chron-early-L80"
   ]
  },
  {
   "id": "F-447",
   "kind": "fact",
   "claim": "On 120 fresh comparison/threshold/superlative probes, 17 (14 %) of the 4B's answers contradicted their own stated facts; the regex guard caught 1",
   "value": "17/120",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "blind probe set (dev now)",
   "set_status": "dev",
   "judge": "reviewer",
   "date": "2026-10-04",
   "source_public": null,
   "caveat": "closed the self-contradiction guard line at v3",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-chron-early-56"
   ]
  },
  {
   "id": "F-448",
   "kind": "event",
   "claim": "ORACLE: reader is not the bottleneck; retrieval is (+0.10 ceiling)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "T-system-L33"
   ]
  },
  {
   "id": "F-449",
   "kind": "event",
   "claim": "Model line closed for the 4B tier (Qwen3.5-4B, Gemma 4 E4B, 9B, 8B)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "E-chron-early-L83"
   ]
  },
  {
   "id": "F-450",
   "kind": "fact",
   "claim": "Bigger closed-book models do not help the 4B tier",
   "value": "Qwen3.5-4B 0.437, Gemma E4B 0.495, Qwen3.5-9B 0.488, Ministral 8B 0.457 vs 0.487",
   "ci": "n 57; 1 seed",
   "n": "57",
   "seeds": "1",
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-05",
   "date_text": "2026-10-05/06",
   "source_public": "boar-research:reports/CLOSED.md:16",
   "caveat": "not detected at MDE 0.055",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-96"
   ]
  },
  {
   "id": "F-451",
   "kind": "fact",
   "claim": "The 4B claim build wins 0 of 51 head-to-heads vs the reference on V-para, both seeds",
   "value": "0/0/51 per seed",
   "ci": "n 51; 2 seeds",
   "n": "51",
   "seeds": "2",
   "set": "V-para, held-out",
   "set_status": "held-out at run time",
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": "boar-research:reports/BAR-TABLE.md:35",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-03"
   ]
  },
  {
   "id": "F-452",
   "kind": "fact",
   "claim": "The 4B claim build's win/tie/loss on v0.2",
   "value": "0/0/110 · 0/1/109",
   "ci": "per seed",
   "n": "110 (inferred from set)",
   "seeds": null,
   "set": "v0.2 (dev)",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": "boar-research:reports/BAR-TABLE.md:53",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-06"
   ]
  },
  {
   "id": "F-453",
   "kind": "fact",
   "claim": "Retrieval failure scorecard on v0.1 gold items",
   "value": "coverage bucket 49 %; recall@5 0.333; gold in prompt 24/78 item-runs",
   "ci": "39 gold items × 2 runs",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-102"
   ]
  },
  {
   "id": "F-454",
   "kind": "fact",
   "claim": "Retrieval captures only part of the oracle gap on v0.1",
   "value": "+0.02 of +0.10",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-103"
   ]
  },
  {
   "id": "F-455",
   "kind": "fact",
   "claim": "Deep model reads gold passages like the 4B",
   "value": "R450 0.634; paired vs 4B +0.029 (−0.004…+0.063)",
   "ci": "1 seed; n 57",
   "n": "57",
   "seeds": "1",
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": "boar-research:reports/CLOSED.md:34",
   "caveat": "deep gain = knowledge, not reading",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-58"
   ]
  },
  {
   "id": "F-456",
   "kind": "event",
   "claim": "Retrieval scorecard: coverage is the largest failure bucket (49 %); recall@5 0.333",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "citation fix: the 49 % coverage bucket is l.6",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "T-results-L38"
   ]
  },
  {
   "id": "F-457",
   "kind": "fact",
   "claim": "L0 retrieval replay on the laptop takes ≈ 8 s with no LLM call; it is the first rung of the funnel.",
   "value": "8 s",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "none",
   "date": "2026-10-07",
   "source_public": null,
   "caveat": "the '8 s' figure is in the workspace map, not at LOOP-PROGRAM l.67 (tool path only)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-method-83"
   ]
  },
  {
   "id": "F-458",
   "kind": "fact",
   "claim": "Routing 4B vs deep per item: in-sample vs out-of-sample",
   "value": "v0.1: 26 % routed, 58 % of gain; V-para: 37 % routed, 43 %/34 % of gain (chance)",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "v0.1 / V-para",
   "set_status": null,
   "judge": "Opus rows",
   "date": "2026-10-07",
   "date_text": "2026-10-07/10",
   "source_public": "boar-research:reports/CLOSED.md:47",
   "caveat": "58 % figure is in-sample only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-61"
   ]
  },
  {
   "id": "F-459",
   "kind": "fact",
   "claim": "Deep tier vs 4B paired gain on V-para",
   "value": "+0.068 floor-adj",
   "ci": "(two-level paired gap file)",
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-08",
   "source_public": "boar-research:reports/BAR-TABLE.md:109",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-12"
   ]
  },
  {
   "id": "F-460",
   "kind": "fact",
   "claim": "H vs the 4B on v0.2",
   "value": "+0.089 raw; +0.111 floor-adj",
   "ci": "raw +0.054…+0.126; floor +0.066…+0.159; 1 seed",
   "n": "110 (inferred from set)",
   "seeds": "1",
   "set": "v0.2 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-09",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-24"
   ]
  },
  {
   "id": "F-461",
   "kind": "fact",
   "claim": "Deep-tier time on desktop.",
   "value": "p50 79 s (V-para; 78.9/79.4 per seed) vs 6 s 4B",
   "ci": "2 seeds",
   "n": "51 (inferred from set)",
   "seeds": "2",
   "set": "V-para",
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "source_public": null,
   "caveat": "Mac mini CPU",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-system-91"
   ]
  },
  {
   "id": "F-462",
   "kind": "event",
   "claim": "GEMMATHINK: Gemma 27 % faster but −0.028 raw → NO",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "date_text": "2026-10-09 06:35",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "T-results-L47"
   ]
  },
  {
   "id": "F-463",
   "kind": "fact",
   "claim": "The 4B's two seeds on V-para give identical ratios (Δ 0.000, −0.023…+0.023).",
   "value": "0.000",
   "ci": "n 51",
   "n": "51",
   "seeds": null,
   "set": "V-para",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-method-36"
   ]
  },
  {
   "id": "F-464",
   "kind": "fact",
   "claim": "The 4B is seed-stable on V-para (identical seeds)",
   "value": "per-item r 0.81; Δ 0.000",
   "ci": "2 seeds",
   "n": "51 (inferred from set)",
   "seeds": "2",
   "set": "V-para",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-22"
   ]
  },
  {
   "id": "F-465",
   "kind": "fact",
   "claim": "Thinking twin of the claim 4B (Qwen3-4B-Thinking-2507 @512) on V-para",
   "value": "raw 0.439; floor-adj 0.293; Δ +0.026 vs K",
   "ci": "Δ CI −0.013…+0.068; 1 seed",
   "n": "51 (inferred from set)",
   "seeds": "1",
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:43",
   "caveat": "×3.7 time (23.6 s); not detected at MDE 0.06",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-34"
   ]
  },
  {
   "id": "F-466",
   "kind": "fact",
   "claim": "4B correctness-only floor-adj on V-para / v0.2",
   "value": "0.427 (0.371–0.482) / 0.595 (0.544–0.646)",
   "ci": "2 seeds",
   "n": "51 (inferred from set)",
   "seeds": "2",
   "set": "V-para / v0.2 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:88,90",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-40"
   ]
  },
  {
   "id": "F-467",
   "kind": "fact",
   "claim": "Shortening references to ≤ 80 words lifts H floor-adj 0.412 → 0.514 and K 0.261 → 0.322",
   "value": "+0.102 (+0.068…+0.144) H; +0.060 K",
   "ci": "1 seed (s42)",
   "n": "51 (inferred from set)",
   "seeds": "1",
   "set": "V-para dev, references altered",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:98",
   "caveat": "≈ 17 % of H's gap is reference length, ≈ 83 % substance; not a bar reading",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-42"
   ]
  },
  {
   "id": "F-468",
   "kind": "fact",
   "claim": "Reference answers are long",
   "value": "p10/p50/p90 = 251/280/318 words",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para",
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:95",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-43"
   ]
  },
  {
   "id": "F-469",
   "kind": "fact",
   "claim": "A Haiku slice sign flipped under Opus once",
   "value": "Haiku +0.005 → Opus −0.027",
   "ci": "n 20 slice",
   "n": "20",
   "seeds": null,
   "set": "v0.2 slice (dev)",
   "set_status": "dev",
   "judge": "both",
   "date": "2026-10-12",
   "source_public": null,
   "caveat": "reason deep/prompt arms still escalate to Opus",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-process-29"
   ]
  },
  {
   "id": "F-470",
   "kind": "fact",
   "claim": "Numbers and dates dominate the wrong-fact loss; declining instead of a wrong answer would lower the score (RATIONALEPASS)",
   "value": "numeric + date 0.51 (V-para) / 0.70 (v0.2) of wrong loss; decline oracle −0.064 / −0.040",
   "ci": "0/16, 0/17 items better",
   "n": null,
   "seeds": null,
   "set": "V-para / v0.2",
   "set_status": null,
   "judge": "jev over Opus rationales",
   "date": "2026-10-12",
   "date_text": "2026-10-12 14:17",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-chron-late-15"
   ]
  },
  {
   "id": "F-472",
   "kind": "fact",
   "claim": "The desktop 4B server path is byte-deterministic (seed 42, one slot, no prompt cache); one divergence in 60 appeared at load average 18.8.",
   "value": "1/60 at high load",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13/14",
   "source_public": "boar-research:reports/RULE-CHANGES.md:34,37",
   "caveat": "identity checks run at load < 4",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-method-81"
   ]
  },
  {
   "id": "F-473",
   "kind": "fact",
   "claim": "A \"noise floor\" of 56/102 changed answers turned out to be a keying artefact: 0/43 same answers when prompt tokens differ, 56/59 when equal.",
   "value": "artefact",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": null,
   "date": "2026-10-13",
   "source_public": "boar-research:reports/RULE-CHANGES.md:34",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-method-82"
   ]
  },
  {
   "id": "F-474",
   "kind": "fact",
   "claim": "An oracle/ceiling cell came first and redirected the line",
   "value": "ceiling +0.095 (+0.070…+0.122) for 1,440 calls and 1.65 MacBook-h",
   "ci": "n 180",
   "n": "180",
   "seeds": null,
   "set": "180 self-tests (dev)",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-13",
   "source_public": null,
   "caveat": "ceiling valid only for questions written from a chunk",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-process-47"
   ]
  },
  {
   "id": "F-475",
   "kind": "fact",
   "claim": "Competitor-question set, our port of their rubric",
   "value": "4B 3.12 (2.58–3.69); deep 3.92 (3.33–4.53); competitor 5.53; Haiku no-web 7.17 (0–10)",
   "ci": "n 72",
   "n": "72",
   "seeds": null,
   "set": "competitor's public questions",
   "set_status": "public dev",
   "judge": "Opus",
   "date": "2026-10-13",
   "date_text": "2026-10-13/14",
   "source_public": null,
   "caveat": "G0 level check failed (their 7.00 → our 5.53); rank-only comparability; obscure-subject collapse 1.38 vs 5.77",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-107"
   ]
  },
  {
   "id": "F-476",
   "kind": "fact",
   "claim": "4B generation on desktop CPU is byte-deterministic",
   "value": "20/20 identical across 5 pairings",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:17",
   "caveat": "citation fix (l.16 is the stack bullet)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-116"
   ]
  },
  {
   "id": "F-477",
   "kind": "event",
   "claim": "Determinism rule (SEEDGEN): pair runs on prompt identity; the desktop 4B path is byte-deterministic",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": "boar-research:reports/RULE-CHANGES.md:34",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "T-method-L58"
   ]
  },
  {
   "id": "F-478",
   "kind": "event",
   "claim": "RIVALSET: on a competitor team's public 72 questions (our port of their rubric) 4B 3.12 vs competitor 5.53 vs Haiku 7.17",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 23:26",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "T-results-L61"
   ]
  },
  {
   "id": "F-479",
   "kind": "event",
   "claim": "RIVALSET: on a competitor's own 72 questions (our port of their rubric) BOAR 4B 3.12 vs competitor 5.53",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 23:28",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "E-chron-late-L61"
   ]
  },
  {
   "id": "F-480",
   "kind": "fact",
   "claim": "Competitor's 72 public questions on our port of their rubric (RIVALSET)",
   "value": "BOAR 4B 3.12 (2.58–3.69); competitor published answers 5.53; Haiku no-web 7.17 (0–10 scale)",
   "ci": "paired −2.41 (−3.08…−1.81); W/T/L 7/7/58",
   "n": null,
   "seeds": null,
   "set": "competitor's public set",
   "set_status": null,
   "judge": "Opus grading",
   "date": "2026-10-13",
   "date_text": "2026-10-13 23:28",
   "source_public": null,
   "caveat": "calibration failed on level (their answers 5.53 on our port vs their own 7.00): never quote on their scale",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-chron-late-43"
   ]
  },
  {
   "id": "F-481",
   "kind": "decision",
   "claim": "SFT of the 4B closed unless a teacher reaches ≥ 50 % faithfulness on distractor-rich targets — why: Haiku teacher 38 %, deep teacher 5 % — outcome: closed (R71)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "D-chron-late-22"
   ]
  },
  {
   "id": "F-482",
   "kind": "fact",
   "claim": "History overhead",
   "value": "≈ 81 tokens per turn; projected 4,096 context at turn ≈ 40–64",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:arms/longsession/LONGSESSION.md:33-34,68",
   "caveat": "81 tok/turn and turn ≈ 40-64 come from the 5-turn study (l.33-34); the 15-turn extension measured ≈ 74 tok/turn and a crossing near turn 50 (l.68)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-76"
   ]
  },
  {
   "id": "F-483",
   "kind": "event",
   "claim": "SFT TEACHER-PROBE NO-GO (deep teacher 6/120 faithful); fine-tuning line closed",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 10:56",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "E-chron-late-L64"
   ]
  },
  {
   "id": "F-484",
   "kind": "fact",
   "claim": "Ternary 2B vs 4B quality",
   "value": "−0.168 (−0.238…−0.105), 2.7× faster",
   "ci": "slice 20",
   "n": "20 (slice)",
   "seeds": null,
   "set": "v0.2 dev slice",
   "set_status": "dev",
   "judge": "Haiku",
   "date": "2026-10-15",
   "source_public": null,
   "caveat": "6/15 unsafe or inverted on model-only health items",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-95"
   ]
  },
  {
   "id": "F-485",
   "kind": "fact",
   "claim": "Cross-family judge (OpenAI gpt-5.6-sol via the Codex CLI) on V-para, 4B: raw 0.497 (0.449-0.547) / floor-adj 0.361 vs Claude 0.413 / 0.261.",
   "value": "+0.084 raw / +0.100 floor",
   "ci": "408 calls total, n 51, 2 seeds",
   "n": "51,",
   "seeds": "2",
   "set": "V-para dev-style",
   "set_status": "dev-style",
   "judge": "cross-family vs Opus",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "post-hoc re-judge",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-method-51"
   ]
  },
  {
   "id": "F-486",
   "kind": "fact",
   "claim": "Cross-family paired deep − 4B on V-para: +0.171 (+0.123…+0.219) vs Claude +0.139; per-item Spearman 0.88-0.92; win-rate 0.00-0.01.",
   "value": "+0.171",
   "ci": "n 51",
   "n": "51",
   "seeds": null,
   "set": "V-para dev-style",
   "set_status": "dev-style",
   "judge": "cross-family",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-method-53"
   ]
  },
  {
   "id": "F-487",
   "kind": "fact",
   "claim": "Registry size",
   "value": "98 rows private (R1–R98); 76 rows public",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": "boar-research:reports/CLOSED.md:6",
   "caveat": "public lags R77–R98",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-110"
   ]
  },
  {
   "id": "F-488",
   "kind": "fact",
   "claim": "Cross-family judge on v0.2: 4B raw 0.617 (0.567-0.666) / floor-adj 0.517 vs Claude 0.512 / 0.388; deep raw 0.714 (0.666-0.759) / floor-adj 0.639 (0.579-0.695) vs 0.608 / 0.508.",
   "value": "≈ +0.10 raw / +0.13 floor",
   "ci": "660 calls, n 110; 4B 2 seeds, deep 1 seed",
   "n": "110",
   "seeds": "2",
   "set": "v0.2, burnt → dev",
   "set_status": "burnt → dev",
   "judge": "cross-family vs Opus",
   "date": "2026-10-17",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-method-54"
   ]
  },
  {
   "id": "F-489",
   "kind": "fact",
   "claim": "On v0.2 the deep − 4B gap is identical under both judge families: +0.097 (+0.053…+0.142) vs +0.096 (+0.062…+0.131); Spearman 0.90-0.97.",
   "value": "Δ 0.001",
   "ci": "n 110",
   "n": "110",
   "seeds": null,
   "set": "v0.2 dev",
   "set_status": "dev",
   "judge": "both",
   "date": "2026-10-17",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-method-55"
   ]
  },
  {
   "id": "F-490",
   "kind": "fact",
   "claim": "Paired deep − 4B on v0.2 is identical under both judges",
   "value": "+0.097 (+0.053…+0.142) vs Claude +0.096 (+0.062…+0.131)",
   "ci": "(+0.053…+0.142) / (+0.062…+0.131)",
   "n": "110 (inferred from set)",
   "seeds": null,
   "set": "v0.2 dev",
   "set_status": "dev",
   "judge": "both",
   "date": "2026-10-17",
   "source_public": null,
   "caveat": "Spearman 0.90–0.97 | on V-para +0.171 vs +0.139",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-results-51",
    "F-chron-late-78"
   ]
  },
  {
   "id": "F-491",
   "kind": "event",
   "claim": "Claim v4.2.3 / v4.2.4: wording as of 2026-10-17 (the posted claim follows the standing at closure); submitted app ports retrieval and prompts, not the health path or deterministic answers",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "date_text": "2026-10-17 20:52–20:59",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "E-chron-late-L100"
   ]
  },
  {
   "id": "F-492",
   "kind": "fact",
   "claim": "The judge system prompt forbids rewarding length, format, confidence or tone and says a correct \"I am not sure\" beats a confident wrong answer.",
   "value": "rule text",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "all",
   "set_status": null,
   "judge": "judge.v1",
   "date": null,
   "source_public": "boar-bench:scorers/vendor/eval-lib/judge-core.mjs:64-75",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-method-06"
   ]
  },
  {
   "id": "F-493",
   "kind": "fact",
   "claim": "The fast-tier reader is Qwen3-4B-Instruct-2507 Q4_K_M.",
   "value": "2,497,281,120 B (≈ 2.5 GB), Apache-2.0",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "date_text": "2026-10",
   "source_public": "boar-research:BOM.md:9",
   "caveat": "HF revision UNKNOWN",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "F-system-01"
   ]
  },
  {
   "id": "F-494",
   "kind": "null-result",
   "claim": "Comparison entity split v1/v2.1 (#3): per-side retrieval for \"X or Y\" → retrieval +, quality Δ −0.0015 (no quality effect on 1.5B)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-early-L281"
   ]
  },
  {
   "id": "F-495",
   "kind": "null-result",
   "claim": "Grounded answer planning (P-2): facets → one cited sentence each → Δ −0.053 (prompt-shaping hurts the 4B)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-early-L288"
   ]
  },
  {
   "id": "F-496",
   "kind": "null-result",
   "claim": "4B self-check (P-3): same-KV check turn → false flags 8/20 (not a usable self-checker)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-early-L289"
   ]
  },
  {
   "id": "F-497",
   "kind": "null-result",
   "claim": "Cross-encoder rerank gate (4B): bge-reranker-v2-m3 → 1 worse, 0 better (the 4B already ignores filtered passages)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-early-L292"
   ]
  },
  {
   "id": "F-498",
   "kind": "null-result",
   "claim": "Retrieval v2 (rv2): windows + dense + relaxed gates → gold 7 → 18/56, quality −0.021 (KILL; retrieval parked)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-early-L297"
   ]
  },
  {
   "id": "F-499",
   "kind": "null-result",
   "claim": "k = 4 → 3 passages (speed): fewer passages → only −19 prompt tokens (token gate failed)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-early-L302"
   ]
  },
  {
   "id": "F-500",
   "kind": "null-result",
   "claim": "Newer 4B / 8–9B readers: Qwen3.5-4B, Gemma 4 E4B, Qwen3.5-9B, Ministral 8B → 0.437 / 0.495 / 0.488 / 0.457 vs 0.487 (none beats Qwen3-4B-2507)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:16-17",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-early-L311"
   ]
  },
  {
   "id": "F-501",
   "kind": "null-result",
   "claim": "History-aware retrieval (FOLLOWQ-G): add history to follow-up queries → A/B 45 % (bar 60 %) (answers not better)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:19",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-early-L314"
   ]
  },
  {
   "id": "F-502",
   "kind": "null-result",
   "claim": "Named-slot retrieval (NAMEDSLOT): one lead passage per named title → 15/60 lists changed (bar 12) (gate FAIL)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:18",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-early-L317"
   ]
  },
  {
   "id": "F-503",
   "kind": "null-result",
   "claim": "Fine-tune data (FTDATA, REJ, HUMANSFT): teacher, rejection, human data → 16 / 11 / 6 of 20 unsupported (bar ≤ 2/20 unreachable; no LoRA)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:27",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-early-L322"
   ]
  },
  {
   "id": "F-504",
   "kind": "null-result",
   "claim": "Hybrid routing by retrieval signal: R2 \"no on-topic source\" → chance-level on V-para (no route-time signal)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-early-L335"
   ]
  },
  {
   "id": "F-505",
   "kind": "null-result",
   "claim": "LOCALGATE: replace the paid jev classifier with the local 4B or Laya → agreement 0.79 / 0.85 (bar 0.90); Laya ≈ always \"false\" (not a drop-in)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-late-L277"
   ]
  },
  {
   "id": "F-506",
   "kind": "null-result",
   "claim": "DENSECHUNK-ST (fusion): dense chunk path fused into the list → displaces kit sources (G2 fail) (harm on the kit)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-late-L278"
   ]
  },
  {
   "id": "F-507",
   "kind": "null-result",
   "claim": "DENSECHUNK-ST2 (extra slots): two extra dense slots → Opus +0.017 (−0.000…+0.035) (not decided at MDE 0.03; format v3 not shipped)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-late-L279"
   ]
  },
  {
   "id": "F-508",
   "kind": "null-result",
   "claim": "SFT4B: LoRA fine-tune of the 4B to read kept passages → teacher faithfulness 38 % (Haiku), 5 % (deep) vs 50 % (no faithful teacher)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-late-L285"
   ]
  },
  {
   "id": "F-509",
   "kind": "null-result",
   "claim": "Bonsai 2 ternary 9B / 4B / 2B: ternary models as smaller/faster tiers → 9B decode 0.50× (speed fail); 4B −0.045 kit; 2B −0.168 and unsafe health (speed or quality)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-late-L290"
   ]
  },
  {
   "id": "F-510",
   "kind": "null-result",
   "claim": "QWRITE / QMERGE: 4B rewrites the search query; merge two query pools → 69 → 67; kept 36 → 42 (bar +10) (reach ≈ 0)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-chron-late-L299"
   ]
  },
  {
   "id": "F-511",
   "kind": "null-result",
   "claim": "Local small model as gate classifier (LOCALGATE): Replace jev with a local 4B classifier → FAIL (Agreement too low; jev stays)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-process-L204"
   ]
  },
  {
   "id": "F-512",
   "kind": "null-result",
   "claim": "R1 Dense ≤ 9B model swap: Qwen3.5-4B, Gemma E4B, Qwen3.5-9B, Ministral 8B closed-book",
   "value": "0.437/0.495/0.488/0.457 vs 0.487",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:16",
   "caveat": "why closed: not detected at MDE 0.055 (57, 1 seed)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L228"
   ]
  },
  {
   "id": "F-513",
   "kind": "null-result",
   "claim": "R3 Named-slot passage ranking: replay, 0 calls",
   "value": "recall@5 0.333 → 0.410 but changed 15/60 lists (bar ≤ 12)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:18",
   "caveat": "why closed: gate FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L230"
   ]
  },
  {
   "id": "F-514",
   "kind": "null-result",
   "claim": "R4 History-aware query: retrieval diagnostic",
   "value": "gains 10/21 follow-ups but changes 11/171 others",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:19",
   "caveat": "why closed: rule not met",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L231"
   ]
  },
  {
   "id": "F-515",
   "kind": "null-result",
   "claim": "R6 Model-written topic notes: 300 notes, replay",
   "value": "note kept 2.7 % (bar 5 %)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:21",
   "caveat": "why closed: gate FAIL step 0",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L233"
   ]
  },
  {
   "id": "F-516",
   "kind": "null-result",
   "claim": "R7 Prompt diet (speed): −126 fixed tokens",
   "value": "quality −0.016; confident-wrong 5 → 9",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:22",
   "caveat": "why closed: non-inferiority failed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L234"
   ]
  },
  {
   "id": "F-517",
   "kind": "null-result",
   "claim": "R8 Completeness instruction: deep reader + gold passages",
   "value": "−0.014 (−0.055…+0.024); confident-wrong 7 → 15",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:23",
   "caveat": "why closed: mechanism raises wrong claims",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L235"
   ]
  },
  {
   "id": "F-518",
   "kind": "null-result",
   "claim": "R10 Thinking 4B, V-para: Qwen3-4B-Thinking @512",
   "value": "+0.026 (−0.013…+0.068), ×3.7 time",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:25",
   "caveat": "why closed: not detected at MDE 0.06",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L237"
   ]
  },
  {
   "id": "F-519",
   "kind": "null-result",
   "claim": "R11 Thinking 4B, v0.2: same, Haiku",
   "value": "+0.009 (−0.015…+0.035)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:26",
   "caveat": "why closed: not detected at MDE 0.025",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L238"
   ]
  },
  {
   "id": "F-520",
   "kind": "null-result",
   "claim": "R12 Fine-tune data (grounded reading SFT): teacher / rejection / human data, strict 20-review",
   "value": "unsupported-claim rate 16/20 → 11/20 → human 6/20; bar ≤ 2/20",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:27",
   "caveat": "why closed: bar unreachable even for human data; no LoRA trained",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L239"
   ]
  },
  {
   "id": "F-521",
   "kind": "null-result",
   "claim": "R13 Multi-pass on the 4B: decompose / research / synthesize",
   "value": "−0.007 (−0.036…+0.023), ×4 cost",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:28",
   "caveat": "why closed: CI upper < +0.03 (borderline)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L240"
   ]
  },
  {
   "id": "F-522",
   "kind": "null-result",
   "claim": "R15 Deep reader advantage: gold passages",
   "value": "+0.029 (−0.004…+0.063)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:34",
   "caveat": "why closed: not detected at MDE 0.055",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L242"
   ]
  },
  {
   "id": "F-523",
   "kind": "null-result",
   "claim": "R20 Budget 512 vs 1024: 2 seeds",
   "value": "+0.011 (−0.010…+0.031), ×0.55 time",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:39",
   "caveat": "why closed: CHEAPER-PARITY → 512 adopted",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L247"
   ]
  },
  {
   "id": "F-524",
   "kind": "null-result",
   "claim": "R21 Budget 256: 2 seeds",
   "value": "−0.028 (−0.056…−0.003)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:40",
   "caveat": "why closed: CHEAPER-LOSS",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L248"
   ]
  },
  {
   "id": "F-525",
   "kind": "null-result",
   "claim": "R22 Budget 3072: Haiku 1 seed",
   "value": "+0.010 (−0.014…+0.041)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:41",
   "caveat": "why closed: not detected; model stops ≈ 1.3k tokens",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L249"
   ]
  },
  {
   "id": "F-526",
   "kind": "null-result",
   "claim": "R23 Deep context 1200 vs 400: slice 20",
   "value": "−0.006 (−0.036…+0.024)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:42",
   "caveat": "why closed: cap barely binds",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L250"
   ]
  },
  {
   "id": "F-527",
   "kind": "null-result",
   "claim": "R24 Long persona cap 200: slice 20",
   "value": "−0.066 (−0.134…+0.014)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:43",
   "caveat": "why closed: truncation",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L251"
   ]
  },
  {
   "id": "F-528",
   "kind": "null-result",
   "claim": "R25 Long persona cap 400: slice 20",
   "value": "−0.013 (−0.069…+0.048)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:44",
   "caveat": "why closed: NO by frozen rule; not detected at 0.08",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L252"
   ]
  },
  {
   "id": "F-529",
   "kind": "null-result",
   "claim": "R26 Vendor sampling presets: slice 20",
   "value": "−0.029 / −0.017",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:45",
   "caveat": "why closed: closed (P borderline)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L253"
   ]
  },
  {
   "id": "F-530",
   "kind": "null-result",
   "claim": "R30 Cross-encoder rerank (deep): replay",
   "value": "gold into top-6 on 2 items (gate 5)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:49",
   "caveat": "why closed: gate FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L257"
   ]
  },
  {
   "id": "F-531",
   "kind": "null-result",
   "claim": "R31 Iterative second retrieval: slice 20, Opus",
   "value": "−0.019 (−0.046…+0.008), +32 % time",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:50",
   "caveat": "R31 'closed' rests on n 20 with CI upper +0.008 under Opus (+0.048 under Haiku); CRITIQUE-2026-10-16.md:58-59 asks that n≈20 closures with CI upper ≥ +0.03 read 'not detected at MDE 0.08'",
   "verification": "OVERCLAIM",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L258"
   ]
  },
  {
   "id": "F-532",
   "kind": "null-result",
   "claim": "R38 Comparison: one pinned source per side: replay",
   "value": "+1/+3 vs bar +5",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:62",
   "caveat": "why closed: step-0 FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L265"
   ]
  },
  {
   "id": "F-533",
   "kind": "null-result",
   "claim": "R40 Anti-hedging prompt: 0 calls",
   "value": "hedged 11 → 11 (items were canned health fallbacks)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:64",
   "caveat": "why closed: instruction never applied",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L267"
   ]
  },
  {
   "id": "F-534",
   "kind": "null-result",
   "claim": "R42 Half-k expert routing: greedy agreement",
   "value": "0.18 / 0.29 (bar 0.80); decode ×1.83",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:66",
   "caveat": "why closed: quality FAIL; parked as speed idea",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L269"
   ]
  },
  {
   "id": "F-535",
   "kind": "null-result",
   "claim": "R43 Multi-sample integration: 3 seeds × 26 wrong items",
   "value": "seed-specific wrong 4 % (bar 50 %)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:67",
   "caveat": "why closed: step-0 FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L270"
   ]
  },
  {
   "id": "F-536",
   "kind": "null-result",
   "claim": "R47 Keep-step grounding relaxations: replay",
   "value": "chunk kept 51–54 vs bar 55",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:71",
   "caveat": "why closed: step-0 FAIL by one item",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L274"
   ]
  },
  {
   "id": "F-537",
   "kind": "null-result",
   "claim": "R50 Country-facts lookup: deterministic",
   "value": "data agreement 84 % (bar 98 %)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:79",
   "caveat": "why closed: gate FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L277"
   ]
  },
  {
   "id": "F-538",
   "kind": "null-result",
   "claim": "R51 Crypto facts tier: coverage gate",
   "value": "2/11 vs ≥ 6",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:80",
   "caveat": "why closed: NO step 1",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L278"
   ]
  },
  {
   "id": "F-539",
   "kind": "null-result",
   "claim": "R52 Crypto EIP selection: 2 seeds",
   "value": "+0.0033 (CI90 −0.005…+0.014)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:81",
   "caveat": "why closed: L1 FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L279"
   ]
  },
  {
   "id": "F-540",
   "kind": "null-result",
   "claim": "R53–R56 Computed verdict (replace / guard / hint / override): deterministic gates + L1",
   "value": "gate gains +6 to +26.5 pp but regressions or L1 Δ ≤ 0",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:82",
   "caveat": "why closed: all FAIL; computed-verdict line closed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L280"
   ]
  },
  {
   "id": "F-541",
   "kind": "null-result",
   "claim": "R57 Verdict guard v4: reviewer gate",
   "value": "1 WRONG of 2 fires; regex catches 1/17 self-contradictions",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:86",
   "caveat": "why closed: FAIL (\"no v5\")",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L281"
   ]
  },
  {
   "id": "F-542",
   "kind": "null-result",
   "claim": "R61–R62 Health follow-up routing fix (+ rescue): deterministic",
   "value": "58 % / 84 % vs bar 90 %",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:90",
   "caveat": "why closed: FAIL; family closed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L283"
   ]
  },
  {
   "id": "F-543",
   "kind": "null-result",
   "claim": "R63 History-aware query for flagged follow-ups: Opus A/B",
   "value": "45 % wins (bar 60 %)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:92",
   "caveat": "why closed: FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L284"
   ]
  },
  {
   "id": "F-544",
   "kind": "null-result",
   "claim": "R65 Guide name in health query: 0 calls",
   "value": "0/18",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:94",
   "caveat": "why closed: step-0 FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L286"
   ]
  },
  {
   "id": "F-545",
   "kind": "null-result",
   "claim": "R66 Speculative decoding on 4B (draft model): 50 prompts",
   "value": "acceptance 31 %; projected saving 8.4 % (bar 20 %)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:95",
   "caveat": "why closed: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L287"
   ]
  },
  {
   "id": "F-546",
   "kind": "null-result",
   "claim": "R67 Key-point enumeration prompt (deep): slice 20 Opus",
   "value": "−0.027 (−0.077…+0.019); correctness −0.11",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:96",
   "caveat": "why closed: NO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L288"
   ]
  },
  {
   "id": "F-547",
   "kind": "null-result",
   "claim": "R69 Keep-step packGround relaxation: replay",
   "value": "14/32 (bar 16)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:98",
   "caveat": "why closed: step-0 FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L290"
   ]
  },
  {
   "id": "F-548",
   "kind": "null-result",
   "claim": "R70 Section selection inside kept articles: replay",
   "value": "3/12 (gate 6/12)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:99",
   "caveat": "why closed: no headroom",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L291"
   ]
  },
  {
   "id": "F-549",
   "kind": "null-result",
   "claim": "R71 Grounded-reading SFT with Haiku teacher: data gates",
   "value": "pass 38 %, 562 kept vs ≥ 1,000",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:100",
   "caveat": "why closed: stage-0 FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L292"
   ]
  },
  {
   "id": "F-550",
   "kind": "null-result",
   "claim": "R72 doc2query (build-time query expansion): 5,975 chunks",
   "value": "reach 69 → 77 vs gate 90",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:101",
   "caveat": "why closed: step-0 FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L293"
   ]
  },
  {
   "id": "F-551",
   "kind": "null-result",
   "claim": "R74 Reranker-chosen append (X3): replay",
   "value": "reach 111/109 vs 117; 21.4 s/item",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:103",
   "caveat": "why closed: step-0 FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L295"
   ]
  },
  {
   "id": "F-552",
   "kind": "null-result",
   "claim": "R76 Multi-turn drift: measurement",
   "value": "no drop to turn 15",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:105",
   "caveat": "why closed: not a lever (measured)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L297"
   ]
  },
  {
   "id": "F-553",
   "kind": "null-result",
   "claim": "R77 Ternary Bonsai 9B (fast tier): speed smoke",
   "value": "decode 0.49× the 4B (bar 0.7×)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: closed at S0 on speed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L298"
   ]
  },
  {
   "id": "F-554",
   "kind": "null-result",
   "claim": "R78 Ternary Bonsai 4B: slice 20 Haiku",
   "value": "−0.045 (−0.109…+0.004)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: NO (parity clause)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L299"
   ]
  },
  {
   "id": "F-555",
   "kind": "null-result",
   "claim": "R82 Ternary 2B as lite model: slice + health layers",
   "value": "−0.168; 6/15 unsafe/inverted model-only health answers",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: not a substitute",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L303"
   ]
  },
  {
   "id": "F-556",
   "kind": "null-result",
   "claim": "R92 Follow-up anchored neighbour fetch: replay",
   "value": "gold kept 1 → 6/20 (gate 7)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: near-miss at reach gate",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L313"
   ]
  },
  {
   "id": "F-557",
   "kind": "null-result",
   "claim": "R96 Two-query pool: replay",
   "value": "kept 36 → 42 (gate +10)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: step-0 FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L317"
   ]
  },
  {
   "id": "F-558",
   "kind": "null-result",
   "claim": "R98 Small model as router: 0 calls",
   "value": "lite ceiling +0.018; deep signal = random at matched share",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: NOT SEEN",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L319"
   ]
  },
  {
   "id": "F-559",
   "kind": "null-result",
   "claim": "rv2 retrieval variant: gold-in-context 7 → 18",
   "value": "Δ −0.021 (CI90 −0.054…+0.009), retrieval 3.6× slower",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: KILL: more gold in context did not translate into quality",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L321"
   ]
  },
  {
   "id": "F-560",
   "kind": "null-result",
   "claim": "k = 4 → 3 (prompt token saving): token gate",
   "value": "−19 tokens (bar −100)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: NO-ADOPT",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L322"
   ]
  },
  {
   "id": "F-561",
   "kind": "null-result",
   "claim": "Native lite claims \"2.5× faster\": recheck",
   "value": "≈ 1.2× at matched quality",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: overstated (not a lever row)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-results-L323"
   ]
  },
  {
   "id": "F-562",
   "kind": "null-result",
   "claim": "Smaller k (4→3) for speed: quick budget knob",
   "value": "−19.3 prompt tokens (bar 100)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: not worth it",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-system-L201"
   ]
  },
  {
   "id": "F-563",
   "kind": "null-result",
   "claim": "Prompt diet: fixed prompt 273 → 147 tokens",
   "value": "quality −0.016, confident-wrong 5 → 9",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md#R7",
   "caveat": "why closed: non-inferiority failed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-system-L202"
   ]
  },
  {
   "id": "F-564",
   "kind": "null-result",
   "claim": "Gemma 4 E4B as reader: swap the 4B",
   "value": "−0.0385 (−0.0725…−0.0058), 1.49× time",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md#R2",
   "caveat": "why closed: worse",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-system-L205"
   ]
  },
  {
   "id": "F-565",
   "kind": "null-result",
   "claim": "Dense ≤ 9B model swap: Qwen3.5-4B/9B, Ministral 8B",
   "value": "0.437–0.495 vs 0.487",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md#R1",
   "caveat": "why closed: not detected at MDE 0.055",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-system-L206"
   ]
  },
  {
   "id": "F-566",
   "kind": "null-result",
   "claim": "Per-item routing 4B vs deep: \"no on-topic source\" rule",
   "value": "V-para capture 34–43 %, ≈ chance",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md#R28",
   "caveat": "why closed: no 0-call signal",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-system-L213"
   ]
  },
  {
   "id": "F-567",
   "kind": "null-result",
   "claim": "Pack routing: vote / 4B classifier",
   "value": "0.25 / 0.58–0.61 accuracy",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md#R48",
   "caveat": "why closed: below 0.85 bar",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-system-L216"
   ]
  },
  {
   "id": "F-568",
   "kind": "null-result",
   "claim": "Ternary models (Bonsai 2): 4B / 2B ternary",
   "value": "lost to our 4B with the kit; 2B −0.168, 6/15 unsafe",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: closed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "walls",
   "aliases": [
    "N-system-L224"
   ]
  },
  {
   "id": "F-569",
   "kind": "event",
   "claim": "First deep-tier test: Qwen3.6-35B-A3B +0.060 over the 4B (first lever above MDE)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-early-L55"
   ]
  },
  {
   "id": "F-570",
   "kind": "fact",
   "claim": "Maple 20B-A1B in the full engine failed: confident-wrong 20/15 vs 10/12, first visible text 1.85× slower",
   "value": "Δ −0.028",
   "ci": "2 seeds",
   "n": null,
   "seeds": "2",
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "dropped the \"extreme MoE reader\" line for the 4B tier",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-early-27"
   ]
  },
  {
   "id": "F-571",
   "kind": "event",
   "claim": "Deep tier (35B MoE) approved; counts only if reported separately from the default",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 ~12:55",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-early-L52"
   ]
  },
  {
   "id": "F-572",
   "kind": "decision",
   "claim": "Park the retrieval line after rv2 failed despite more gold in context — why: gold-sentence-in-context did not translate into judged quality — outcome: retrieval reopened later only for the deep tier",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "D-chron-early-19"
   ]
  },
  {
   "id": "F-573",
   "kind": "event",
   "claim": "Token budget: work pauses; deep tier paused at D1",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "date_text": "2026-10-03 evening",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-early-L77"
   ]
  },
  {
   "id": "F-574",
   "kind": "event",
   "claim": "Direction: 4B default tier only; deep tier parked; no team chat agents, coordinator + subagents",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-early-L79"
   ]
  },
  {
   "id": "F-575",
   "kind": "fact",
   "claim": "Gemma 4 E4B inside the full engine lost to Qwen3-4B",
   "value": "Δ −0.0385",
   "ci": "CI90 −0.0725…−0.0058, 2 seeds",
   "n": null,
   "seeds": "2",
   "set": "v0.1 EN dev",
   "set_status": "dev",
   "judge": "Opus-med",
   "date": "2026-10-05",
   "source_public": "boar-research:reports/CLOSED.md:17",
   "caveat": "engine forced hidden thinking on Gemma",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-early-59"
   ]
  },
  {
   "id": "F-576",
   "kind": "decision",
   "claim": "Stop answer-identical 4B arms; reopen the deep tier as the primary quality lever — why: 15 arms, 0 ratio movement — outcome: deep-tier line ran 10-07 → 10-10",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "D-chron-early-24"
   ]
  },
  {
   "id": "F-577",
   "kind": "event",
   "claim": "Directional day: DEEPREAD, COMPLETE, NAMEDSLOT, FACTS; \"reader-capability gap\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-early-L95"
   ]
  },
  {
   "id": "F-578",
   "kind": "fact",
   "claim": "Deep 35B reading the same gold passages as the 4B scored only slightly higher: the reader converts like the 4B; engine-level +0.06 is knowledge",
   "value": "0.634 vs 0.606; Δ +0.029",
   "ci": "(−0.004…+0.063), 57 items, 1 seed",
   "n": null,
   "seeds": "1",
   "set": "v0.1 EN dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": "boar-research:reports/CLOSED.md:34",
   "caveat": "\"not detected at MDE 0.055\"",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-early-75"
   ]
  },
  {
   "id": "F-579",
   "kind": "fact",
   "claim": "A completeness instruction made the 35B longer but less correct",
   "value": "correctness 3.89 → 3.10; CW 7 → 15; ratio Δ −0.014",
   "ci": "(−0.055…+0.024), 1 seed",
   "n": null,
   "seeds": "1",
   "set": "v0.1 EN dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-06",
   "source_public": "boar-research:reports/CLOSED.md:23",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-early-76"
   ]
  },
  {
   "id": "F-580",
   "kind": "event",
   "claim": "DEEPREAD: deep model reads gold passages like the 4B (+0.029, CI −0.004…+0.063) → its gain is knowledge",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "T-results-L36"
   ]
  },
  {
   "id": "F-581",
   "kind": "fact",
   "claim": "Closed-book MoE plateau: Gemma 4 26B-A4B 0.613, Qwen3.5-35B 0.603, Qwen3.6-35B 0.596 vs 4B 0.487",
   "value": "Δ vs 4B +0.126 / +0.116 / +0.108",
   "ci": "57 items, 1 seed, 342 calls",
   "n": null,
   "seeds": "1",
   "set": "v0.1 EN dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-07",
   "source_public": "boar-research:reports/CLOSED.md:36-37",
   "caveat": "none ≥ +0.03 over deep35",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-early-81"
   ]
  },
  {
   "id": "F-582",
   "kind": "fact",
   "claim": "Real retrieval does not hurt the 35B; the closed-book vs engine gap is answer length/protocol",
   "value": "Δ −0.031 (−0.072…+0.011)",
   "ci": "57 items",
   "n": null,
   "seeds": null,
   "set": "v0.1 EN dev",
   "set_status": "dev",
   "judge": "existing rows",
   "date": "2026-10-07",
   "source_public": "boar-research:reports/CLOSED.md:33",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-early-82"
   ]
  },
  {
   "id": "F-583",
   "kind": "fact",
   "claim": "The Gemma judge preserved the ranking: deep35 − 4B +0.026 vs Opus +0.029.",
   "value": "+0.026 vs +0.029",
   "ci": "n 57",
   "n": "57",
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "both",
   "date": "2026-10-07",
   "source_public": null,
   "caveat": "neither CI excludes 0",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-method-47"
   ]
  },
  {
   "id": "F-584",
   "kind": "fact",
   "claim": "Local open judge (Gemma 4 26B) scores +0.11 above Opus but keeps the ranking",
   "value": "deep35 − 4B +0.026 vs Opus +0.029; κ 0.32/0.55",
   "ci": "228 calls",
   "n": null,
   "seeds": null,
   "set": "v0.1 public dev",
   "set_status": "public dev",
   "judge": "local Gemma",
   "date": "2026-10-07",
   "source_public": "boar-research:README.md:77",
   "caveat": "ranking-only judge",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-results-54"
   ]
  },
  {
   "id": "F-585",
   "kind": "fact",
   "claim": "Fast-tier constants in cut 7.",
   "value": "retrieve k 6, context 1,200 tokens (lookup 3/450; other 4/700 quick budget), fast thinking budget 256 (unused by the instruct 4B)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": "instant final confidence 0.75 (context.ts:139)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-system-23"
   ]
  },
  {
   "id": "F-586",
   "kind": "event",
   "claim": "MOESCREEN: 25–35B MoE closed-book plateau ≈ 0.60 vs 4B 0.487",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "T-results-L39"
   ]
  },
  {
   "id": "F-587",
   "kind": "event",
   "claim": "MoE screen: 25–35B MoEs form one closed-book plateau ≈ 0.60",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "date_text": "2026-10-07 00:11–03:55",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-early-L98"
   ]
  },
  {
   "id": "F-588",
   "kind": "decision",
   "claim": "Autonomous mandate: keep running rounds without asking; stop only for irreversible/expensive items — why: throughput — outcome: deep-tier rounds ran unattended",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "date_text": "2026-10-07 14:26",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "D-chron-early-28"
   ]
  },
  {
   "id": "F-589",
   "kind": "event",
   "claim": "Deep tier on V-para / v0.2 (DEEPVPARA): floor-adj 0.329 / 0.457",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-08",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-early-L105"
   ]
  },
  {
   "id": "F-590",
   "kind": "event",
   "claim": "Thinking mode on the deep tier: V-para floor-adj 0.422",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-08",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-early-L106"
   ]
  },
  {
   "id": "F-591",
   "kind": "event",
   "claim": "Opus judge sharding for claim numbers; 20-item slice before any full deep-tier run",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-08",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-process-L42"
   ]
  },
  {
   "id": "F-592",
   "kind": "fact",
   "claim": "Thinking mode (budget 1024) added the largest single gain: V-para floor-adj +0.093",
   "value": "0.422 (0.385–0.459); Δ +0.093 (+0.050…+0.137)",
   "ci": "2 seeds, 306 calls",
   "n": null,
   "seeds": "2",
   "set": "V-para dev-style",
   "set_status": "dev-style",
   "judge": "Opus",
   "date": "2026-10-08",
   "source_public": "boar-research:reports/BAR-TABLE.md:38,106",
   "caveat": "×4.2 time (p50 ~143 s desktop)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-early-92"
   ]
  },
  {
   "id": "F-593",
   "kind": "fact",
   "claim": "Budget curve (V-para floor-adj): 0 → 256 → 512 → 1024 → 3072",
   "value": "0.329 → 0.387 → 0.436 → 0.422 → no further gain",
   "ci": "2 seeds each (3072: 1 seed Haiku, +0.010)",
   "n": "51 (inferred from set)",
   "seeds": "2",
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "Opus (3072 Haiku)",
   "date": "2026-10-08",
   "date_text": "2026-10-08/09",
   "source_public": "boar-research:reports/BAR-TABLE.md:37-41; boar-research:reports/CLOSED.md:41",
   "caveat": "knee at 512; Qwen stops thinking ≈ 1.3k tokens",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-results-29"
   ]
  },
  {
   "id": "F-594",
   "kind": "event",
   "claim": "SLICE20: every deep-tier / held-out-style arm starts on a 20-item slice",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-08",
   "source_public": "boar-research:reports/RULE-CHANGES.md:13",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "T-method-L46"
   ]
  },
  {
   "id": "F-595",
   "kind": "event",
   "claim": "Deep tier thinking budget curve; 512 adopted (DT512)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-08",
   "date_text": "2026-10-08/09",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "T-system-L50"
   ]
  },
  {
   "id": "F-596",
   "kind": "event",
   "claim": "DEEPVPARA: deep tier on V-para floor-adj 0.329 vs 4B 0.261 (two seeds, Opus)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-08",
   "date_text": "2026-10-08 15:57",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "T-results-L44"
   ]
  },
  {
   "id": "F-597",
   "kind": "event",
   "claim": "DEEPTHINK: deep + thinking 1024 = 0.422 floor-adj; +0.093 over no thinking",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-08",
   "date_text": "2026-10-08 20:20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "T-results-L45"
   ]
  },
  {
   "id": "F-598",
   "kind": "decision",
   "claim": "Recommend thinking budget 512 for the deep tier — why: CHEAPER-PARITY with 1024 at 0.55× time — outcome: becomes the \"recommended deep configuration\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "D-chron-early-31"
   ]
  },
  {
   "id": "F-599",
   "kind": "decision",
   "claim": "Recommend deep tier = cut 6 + Qwen3.6-35B-A3B + thinking budget 512 (why: DT512 CHEAPER-PARITY: = 1024 quality at 0.55× time)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "source_public": null,
   "caveat": "outcome: Became row H everywhere; selected best of ≥ 8 deep configs on V-para (winner's-curse ≈ 0.01–0.02)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "D-results-06"
   ]
  },
  {
   "id": "F-600",
   "kind": "event",
   "claim": "Gemma 4 26B + thinking: faster, not at parity → NO",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-early-L108"
   ]
  },
  {
   "id": "F-601",
   "kind": "event",
   "claim": "v0.2 deep + thinking two seeds: floor-adj 0.514 (CI straddles 0.50), win 0/110",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-early-L110"
   ]
  },
  {
   "id": "F-602",
   "kind": "fact",
   "claim": "Deep + thinking 1024 on v0.2 (two seeds): first point above 0.50 floor-adj, CI straddles",
   "value": "0.514 (0.475–0.554); raw 0.612",
   "ci": "2 seeds",
   "n": null,
   "seeds": "2",
   "set": "v0.2 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-09",
   "source_public": "boar-research:reports/BAR-TABLE.md:56",
   "caveat": "win 0/110",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-early-94"
   ]
  },
  {
   "id": "F-603",
   "kind": "fact",
   "claim": "Gemma 4 26B + thinking was 27 % faster but below Qwen + thinking",
   "value": "floor-adj 0.388 vs 0.422; raw Δ −0.028 (−0.057…+0.003)",
   "ci": "2 seeds",
   "n": null,
   "seeds": "2",
   "set": "V-para dev-style",
   "set_status": "dev-style",
   "judge": "Opus",
   "date": "2026-10-09",
   "source_public": "boar-research:reports/CLOSED.md:38",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-early-95"
   ]
  },
  {
   "id": "F-604",
   "kind": "fact",
   "claim": "Thinking budget 512 = 1024 in quality at 0.55× the time",
   "value": "floor-adj 0.436 (0.401-0.472); raw 0.553 (0.524-0.581); Δ vs 1024 raw +0.011 (-0.010…+0.031), floor-adj +0.014 (-0.012…+0.039)",
   "ci": "2 seeds",
   "n": null,
   "seeds": "2",
   "set": "V-para dev-style",
   "set_status": "dev-style",
   "judge": "Opus",
   "date": "2026-10-09",
   "source_public": "boar-research:reports/BAR-TABLE.md:39,107",
   "caveat": "p50 79 s vs 143 s desktop | Verification note: The quoted Δ is the RAW delta; floor-adj Δ is from BAR-TABLE:107. Minor public-source conflict: BAR-TABLE §1a gives raw 0.553, BAR-TABLE §2 and QUEUE:842 give 0.552. Use 0.553 (two-seed L2 aggregate) consistently.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-early-96"
   ]
  },
  {
   "id": "F-605",
   "kind": "fact",
   "claim": "Haiku L1 stopped the v0.2 thinking arm at +0.027 (vs +0.03 clause); Opus later measured +0.045 (+0.022…+0.070)",
   "value": "+0.027 Haiku / +0.045 Opus",
   "ci": "Haiku s42: +0.027 (+0.002…+0.054), n 110; Opus two seeds two-level: +0.045 (+0.022…+0.070); Opus same-seed s42 comparison: +0.048",
   "n": "110",
   "seeds": "Haiku 1; Opus 2",
   "set": "v0.2 dev",
   "set_status": "dev",
   "judge": "Haiku L1 / Opus",
   "date": "2026-10-09",
   "source_public": "boar-research:reports/CLOSED.md:11; boar-research:method/L1-PROTOCOL.md:20",
   "caveat": "+0.045 is the two-seed figure (not '1 seed'); the near-miss rule text quotes the same-seed +0.048. Both are correct in context",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-results-17"
   ]
  },
  {
   "id": "F-606",
   "kind": "fact",
   "claim": "Budget 256 vs 1024 / vs 512",
   "value": "−0.028 / −0.039 raw",
   "ci": "−0.056…−0.003 / −0.066…−0.015; 2 seeds",
   "n": "51 (inferred from set)",
   "seeds": "2",
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-09",
   "source_public": "boar-research:reports/CLOSED.md:40",
   "caveat": "keeps ≈ 63 % of the thinking gain at 40 % of the time",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-results-28"
   ]
  },
  {
   "id": "F-607",
   "kind": "fact",
   "claim": "Thinking budget hits: at 1024, 58/91 generating v0.2 items and 36/51 V-para; at 512, 83/110 v0.2 and 46/51 V-para",
   "value": "1024: 58/91 generating v0.2 items (s42) and 36/51 V-para (s42); 512: 83/110 v0.2 (s42) and 46/51 (s42) / 42/51 (s43) V-para",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "v0.2 / V-para dev",
   "set_status": "dev",
   "judge": null,
   "date": "2026-10-09",
   "source_public": null,
   "caveat": "the 36/51 figure was uncited (it is DEEPTHINK.md l.16); counts are per seed",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-results-31"
   ]
  },
  {
   "id": "F-608",
   "kind": "fact",
   "claim": "Gemma + thinking vs Qwen + thinking",
   "value": "−0.028 raw",
   "ci": "−0.057…+0.003; 2 seeds",
   "n": "51 (inferred from set)",
   "seeds": "2",
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-09",
   "source_public": "boar-research:reports/CLOSED.md:38",
   "caveat": "parity-faster clause failed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-results-33"
   ]
  },
  {
   "id": "F-609",
   "kind": "event",
   "claim": "DEEPTHINK-V02 L1 NO at Haiku +0.027 (later Opus addendum: v0.2 floor-adj 0.514)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "date_text": "2026-10-09 02:32",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "T-results-L46"
   ]
  },
  {
   "id": "F-610",
   "kind": "event",
   "claim": "DT512: thinking budget 512 = 1024 quality at 0.55× time → recommended config",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "date_text": "2026-10-09 18:21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "T-results-L48"
   ]
  },
  {
   "id": "F-611",
   "kind": "event",
   "claim": "DT256: budget 256 loses −0.028 vs 1024 (CHEAPER-LOSS); knee at 512",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "date_text": "2026-10-09 20:45",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "T-results-L49"
   ]
  },
  {
   "id": "F-612",
   "kind": "event",
   "claim": "DT512-V02: H on v0.2 floor-adj 0.508 (0.467–0.550), first point above 0.50, CI straddles",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "date_text": "2026-10-09 23:14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "T-results-L50"
   ]
  },
  {
   "id": "F-613",
   "kind": "decision",
   "claim": "Near-miss rule: a Haiku Δ within 0.01 of a threshold is re-judged with Opus before deciding (4B arms: slice Δ ≥ +0.02) (why: DEEPTHINK-V02 stopped at +0.027 under Haiku, Opus later +0.048)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": "boar-research:method/L1-PROTOCOL.md:25-26",
   "caveat": "outcome: fired twice on THINK4B and changed the course both times",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "D-method-17"
   ]
  },
  {
   "id": "F-614",
   "kind": "decision",
   "claim": "Declare the deep-tier line closed on desktop and the 4B side closed (why: model, parity, thinking, budget, persona/context, routing all measured)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": "outcome: Research moved to Knowledge Packs (10-12)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "D-results-07"
   ]
  },
  {
   "id": "F-615",
   "kind": "event",
   "claim": "Best-of-2, sampling presets, context size, corpus size, thinking 4B: all NO",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-early-L111"
   ]
  },
  {
   "id": "F-616",
   "kind": "fact",
   "claim": "Routing only \"no on-topic source\" items to the deep tier: chance-level on V-para, informative on v0.2",
   "value": "V-para +0.052 raw at 37 % routed; v0.2 +0.042 at 30 %",
   "ci": "2 seeds / 1 seed",
   "n": null,
   "seeds": "2",
   "set": "V-para dev-style; v0.2 dev",
   "set_status": "dev-style",
   "judge": "existing Opus rows",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:45,61; boar-research:reports/CLOSED.md:47",
   "caveat": "descriptive",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-early-104"
   ]
  },
  {
   "id": "F-617",
   "kind": "fact",
   "claim": "Process review: 65/65 criterion shas verify; DEEPTHINK yielded 0.30 floor-adj points per 1k judge calls (×10 the deep-tier entry arm)",
   "value": "65/65",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:106-109",
   "caveat": "Verification note: Line refs are :13 (65/65) and :18 (DEEPTHINK). Unit: 0.30 floor-adj per 1k calls = 30.5 \"points\" (100×Δ) per 1k calls in public BAR-TABLE:106.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-early-106"
   ]
  },
  {
   "id": "F-618",
   "kind": "fact",
   "claim": "Per-item seed agreement: deep thinking tier r 0.5-0.7 on V-para vs 0.81 for the 4B; on v0.2 the deep tier r 0.90.",
   "value": "r",
   "ci": "n 51 / 110",
   "n": "51",
   "seeds": null,
   "set": "V-para / v0.2 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-method-35"
   ]
  },
  {
   "id": "F-619",
   "kind": "fact",
   "claim": "Haiku tracked Opus's sign 5/5 and size within ±0.02 in 4/5 deep-tier arms; one miss stopped an arm (Haiku +0.027 vs Opus +0.048).",
   "value": "5/5 sign",
   "ci": "5 arms",
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para / v0.2 dev",
   "set_status": "dev",
   "judge": "both",
   "date": "2026-10-10",
   "source_public": "boar-research:method/L1-PROTOCOL.md:17-25",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-method-42"
   ]
  },
  {
   "id": "F-620",
   "kind": "fact",
   "claim": "Pipeline yield: the adopted deep-tier arms gave 16.4 V-para floor-adj points per 1,000 judge calls; one later cycle gave 0 adopted points for 1,804 calls.",
   "value": "16.4 / 0",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para dev",
   "set_status": "dev",
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10/12",
   "source_public": "boar-research:reports/BAR-TABLE.md:108,141",
   "caveat": "citation fix: 0 points per 1,804 calls is l.141 (l.124 is a header)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-method-75"
   ]
  },
  {
   "id": "F-621",
   "kind": "fact",
   "claim": "Highest-yield arm per judge call (thinking on the deep tier)",
   "value": "+0.093 floor-adj for 306 calls ≈ 0.30 points / 1k calls",
   "ci": "2 seeds",
   "n": null,
   "seeds": "2",
   "set": "V-para (dev-style)",
   "set_status": "dev-style",
   "judge": "Opus",
   "date": "2026-10-10",
   "source_public": null,
   "caveat": "desktop only | Verification note: Line ref is :18. Public BAR-TABLE:106 expresses the same as 30.5 \"points\" (100×Δ) per 1k calls.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-process-52"
   ]
  },
  {
   "id": "F-622",
   "kind": "fact",
   "claim": "Deep-tier constants.",
   "value": "retrieve k 6, context 400 tokens, max 200 answer tokens; thinking budget 1024 in code, 512 recommended",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-system-24"
   ]
  },
  {
   "id": "F-623",
   "kind": "event",
   "claim": "THINK4B: thinking twin of the 4B +0.026 (CI crosses 0), ×3.7 time → NO",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "T-results-L51"
   ]
  },
  {
   "id": "F-624",
   "kind": "event",
   "claim": "Desktop queue empty: 4B side and deep-tier line both closed",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "T-results-L54"
   ]
  },
  {
   "id": "F-625",
   "kind": "event",
   "claim": "DEEPRERANK, MP4B, DEEPITER, DEEPCOV-FB closed (retrieval-side levers on the deep tier)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 22:32–23:43",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-late-L27"
   ]
  },
  {
   "id": "F-626",
   "kind": "fact",
   "claim": "Retrieval kept nothing on half the items",
   "value": "29/51 V-para, 54/110 v0.2 = 55 % / 50 % of the deep tier's loss",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "same",
   "set_status": null,
   "judge": "same",
   "date": "2026-10-11",
   "date_text": "10-11",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-late-10"
   ]
  },
  {
   "id": "F-627",
   "kind": "fact",
   "claim": "The deep tier's gain over the 4B came from repairing wrong answers, not from absent knowledge",
   "value": "+0.33 / +0.19 on 4B-wrong items vs +0.02 / +0.06 on absent",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "V-para / v0.2",
   "set_status": null,
   "judge": "Opus + jev",
   "date": "2026-10-11",
   "date_text": "10-11",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-late-11"
   ]
  },
  {
   "id": "F-628",
   "kind": "fact",
   "claim": "Common-mode CI widens deep-tier intervals ≈ 1.2-1.3×; e.g. deep V-para floor-adj 0.401-0.472 → 0.387-0.486.",
   "value": "×1.2-1.3",
   "ci": "2 seeds",
   "n": "51 (inferred from set)",
   "seeds": "2",
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-11",
   "source_public": "boar-research:reports/BAR-TABLE.md:122-131",
   "caveat": "citation fix (l.137-143 are the 10-12 addendum)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-method-41"
   ]
  },
  {
   "id": "F-629",
   "kind": "event",
   "claim": "Loop tick 1: LOSSMAP measures where the deep tier loses points",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "date_text": "2026-10-11 12:10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-late-L31"
   ]
  },
  {
   "id": "F-630",
   "kind": "fact",
   "claim": "Where the deep tier loses points (LOSSMAP)",
   "value": "absent knowledge 0.59 (V-para) / 0.55 (v0.2) of loss; wrong fact 0.40 / 0.21",
   "ci": "51 / 110 items; 1 seed",
   "n": null,
   "seeds": "1",
   "set": "V-para dev-style + v0.2 dev",
   "set_status": "dev-style",
   "judge": "Opus rationales classified by jev (a calibrated classifier)",
   "date": "2026-10-11",
   "date_text": "2026-10-11 12:10",
   "source_public": null,
   "caveat": "label model unvalidated against humans (CRITIQUE C3)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-late-09"
   ]
  },
  {
   "id": "F-631",
   "kind": "fact",
   "claim": "n needed to resolve Δ 0.03 (80 % power): 4B arms ≈ 76-80 items one seed; deep/thinking arms ≈ 213 one seed or ≈ 119 two seeds.",
   "value": "n",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para / v0.2",
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": "boar-bench:bench/v0.3/plan/power-statement.md:6-12",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-method-67"
   ]
  },
  {
   "id": "F-632",
   "kind": "fact",
   "claim": "3-bit quant of the deep model (UD-IQ3_XXS) does not move the deep tier",
   "value": "V-para +0.004 (−0.024…+0.033); v0.2 +0.008 (−0.010…+0.028)",
   "ci": "V-para 2 seeds; v0.2 1 seed",
   "n": "51 (inferred from set)",
   "seeds": "2",
   "set": "V-para / v0.2 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-12",
   "source_public": "boar-research:reports/CLOSED.md:61",
   "caveat": "repairs \"wrong\" bucket on v0.2 (+0.087)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-results-59"
   ]
  },
  {
   "id": "F-633",
   "kind": "event",
   "claim": "DEEPCOVER NO (correctness −0.11), VITALADD NO at slice",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 00:34–00:39",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-late-L34"
   ]
  },
  {
   "id": "F-634",
   "kind": "event",
   "claim": "QUANTCEIL (3-bit vs 2-bit quant of the 35B): slice positive, full v0.2 +0.008, V-para two seeds +0.004 → not detected",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 01:28–07:55",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-late-L35"
   ]
  },
  {
   "id": "F-635",
   "kind": "event",
   "claim": "SYNTHDEEP, WIKISECTION, DEEPHEDGE closed at 0-call gates",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 08:07–09:32",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-late-L36"
   ]
  },
  {
   "id": "F-636",
   "kind": "event",
   "claim": "Sequencing incident: the 35B Q2 model was deleted for QUANTCEIL while DEEPHEDGE needed it",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 08:21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-late-L37"
   ]
  },
  {
   "id": "F-637",
   "kind": "event",
   "claim": "Model swap for one arm deletes the 35B model another arm still needs (second time) → rule 22",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 08:21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-process-L50"
   ]
  },
  {
   "id": "F-638",
   "kind": "fact",
   "claim": "Wrong facts of the deep tier are stable across seeds (SEEDSPEC)",
   "value": "seed-specific 4 % (35 % counting mixed) vs ≥ 50 % bar",
   "ci": "26 wrong items, 3 seeds",
   "n": null,
   "seeds": "3",
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "jev",
   "date": "2026-10-12",
   "date_text": "2026-10-12 14:32",
   "source_public": null,
   "caveat": "closes multi-sample integration",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-late-14"
   ]
  },
  {
   "id": "F-639",
   "kind": "fact",
   "claim": "4B generation on the desktop CPU path is byte-deterministic.",
   "value": "20/20 identical across 5 pairings (one 8 h apart)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:17",
   "caveat": "deep tier not probed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-system-74"
   ]
  },
  {
   "id": "F-640",
   "kind": "fact",
   "claim": "Reader-side fine-tuning teacher faithfulness (SFT4B)",
   "value": "Haiku teacher 38–45 %; deep-tier teacher 6/120 (5 %) vs 50 % bar",
   "ci": "120 contexts",
   "n": null,
   "seeds": null,
   "set": "distractor-rich engine renderings",
   "set_status": null,
   "judge": "jev",
   "date": "2026-10-13",
   "date_text": "2026-10-13 22:28 / 10-14 10:56",
   "source_public": null,
   "caveat": "45/120 deep answers empty (shared token budget)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-late-45"
   ]
  },
  {
   "id": "F-641",
   "kind": "fact",
   "claim": "A teacher-model probe ran 10.1 mini-hours without a smoke test; 45/120 answers were empty because thinking and answer shared one token budget",
   "value": "10.1 h; 45/120 empty; 6/120 passing",
   "ci": "n 120",
   "n": "120",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "jev check",
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "a 10-item smoke (≈ 54 min) would have shown it",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-process-23"
   ]
  },
  {
   "id": "F-642",
   "kind": "fact",
   "claim": "A simple ceiling formula predicted four retrieval arms' Opus results within 0.004",
   "value": "ceiling = new facts present / n × 0.18; e.g. +0.067 vs Opus +0.066",
   "ci": "4 arms",
   "n": null,
   "seeds": null,
   "set": "180 self-tests (dev)",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "0.18 was calibrated on one of the four (circular); does not transfer to the deep tier",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-process-46"
   ]
  },
  {
   "id": "F-643",
   "kind": "event",
   "claim": "RIVALSET-DEEP: deep tier 3.92 on the competitor port, gap is obscure-subject coverage",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 17:06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-late-L67"
   ]
  },
  {
   "id": "F-644",
   "kind": "fact",
   "claim": "Same set with the deep tier (RIVALSET-DEEP)",
   "value": "3.92 (3.33–4.53); − competitor −1.60 (−2.31…−0.94); obscure subjects 1.92 vs 5.77",
   "ci": "n 72",
   "n": "72",
   "seeds": null,
   "set": "competitor's public set",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-14",
   "date_text": "2026-10-14 17:06",
   "source_public": null,
   "caveat": "gap = corpus coverage (their ≈ 20 GB vs kit 2.3 GB)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-chron-late-44"
   ]
  },
  {
   "id": "F-645",
   "kind": "decision",
   "claim": "Lite tier: thinking stays OFF; 1.7B Q4_K_M chosen over ternary 2B and Q8_0 (why: LITE2 2B −0.057; LITE-THINK flat/negative; QUANTLITE flat and slower)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15/16",
   "source_public": null,
   "caveat": "outcome: Lite tier remains a product idea, not a claim",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "D-results-15"
   ]
  },
  {
   "id": "F-646",
   "kind": "fact",
   "claim": "Cross-family judge level offset on V-para",
   "value": "Claude stricter by ≈ +0.10; 4B floor-adj 0.361 (Codex) vs 0.261 (Claude); deep+think 512 floor-adj 0.577 (0.519-0.632) vs 0.436",
   "ci": "n 51, 2 seeds, 408 calls, 0 generation",
   "n": "51,",
   "seeds": "2",
   "set": "V-para (dev-style for deep; held-out for 4B)",
   "set_status": "held-out (at run time)",
   "judge": "cross-family Codex vs Claude Opus",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "win-rate 0.00-0.01 under both; ranking unchanged (Spearman 0.88-0.92)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-process-65"
   ]
  },
  {
   "id": "F-647",
   "kind": "event",
   "claim": "Lite line closed: readers, quant, reach budget, follow-up, query rewrite, thinking",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 10:05–15:06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-chron-late-L91"
   ]
  },
  {
   "id": "F-648",
   "kind": "fact",
   "claim": "Cross-family offset replicates on v0.2",
   "value": "4B floor-adj 0.517 vs 0.388; deep+think 0.639 vs 0.508; H − K +0.097 vs +0.096",
   "ci": "n 110, 660 calls",
   "n": "110,",
   "seeds": "4B 2 seeds; deep+think 512 1 seed",
   "set": "v0.2 (burnt → dev)",
   "set_status": "burnt → dev",
   "judge": "Codex vs Claude Opus",
   "date": "2026-10-17",
   "source_public": null,
   "caveat": "dev data | Verification note: Seeds added: 4B two seeds, deep one seed (QUEUE:1280). Dev data; post-hoc, not pre-registered.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-process-66"
   ]
  },
  {
   "id": "F-649",
   "kind": "fact",
   "claim": "Blinding strips think blocks, trailing \"Sources:\" lists, URLs, markdown links, [n] markers and bold before judging.",
   "value": "1 function",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "all",
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-bench:scorers/vendor/eval-lib/judge-core.mjs:5-20",
   "caveat": "stripping sources also removes web provenance signals from the reference",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-method-07"
   ]
  },
  {
   "id": "F-650",
   "kind": "fact",
   "claim": "The deep-tier model is Qwen3.6-35B-A3B UD-Q2_K_XL.",
   "value": "12,290,628,576 B, Apache-2.0",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "date_text": "2026-10",
   "source_public": "boar-research:BOM.md:11",
   "caveat": "desktop only so far",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-system-03"
   ]
  },
  {
   "id": "F-651",
   "kind": "null-result",
   "claim": "Gemma 4 E4B as 4B-tier reader (#2): swap reader, 2 seeds → Δ −0.016; more confident-wrong; hidden thinking slow (criterion (1) failed; inside noise)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-early-L280"
   ]
  },
  {
   "id": "F-652",
   "kind": "null-result",
   "claim": "Maple 20B-A1B fast reader (#15): ternary MoE in engine → CW 20/15 vs 10/12; 1.85× slower first text (FAIL → extreme-MoE reader line dropped)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-early-L287"
   ]
  },
  {
   "id": "F-653",
   "kind": "null-result",
   "claim": "Completeness prompt on the 35B (COMPLETE): 120–220 words → correctness −0.79, CW 7 → 15 (prompting buys completeness with errors)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:23",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-early-L318"
   ]
  },
  {
   "id": "F-654",
   "kind": "null-result",
   "claim": "MoE closed-book screen: Qwen3.5-35B, Gemma 4 26B vs deep35 → +0.007 / +0.018 (below +0.03 bar; plateau ≈ 0.60)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:36-37",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-early-L321"
   ]
  },
  {
   "id": "F-655",
   "kind": "null-result",
   "claim": "Deep thinking 3072 (DEEPTHINK-2): bigger budget → +0.010 (Haiku) (model stops ≈ 1.3k tokens)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-early-L325"
   ]
  },
  {
   "id": "F-656",
   "kind": "null-result",
   "claim": "Gemma 4 26B + thinking (GEMMATHINK): faster deep model → −0.028 raw (parity clause failed)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-early-L326"
   ]
  },
  {
   "id": "F-657",
   "kind": "null-result",
   "claim": "Long persona on deep tier (DEEPFULL): \"detailed\" persona → −0.066 / −0.013 (NO at slice)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-early-L327"
   ]
  },
  {
   "id": "F-658",
   "kind": "null-result",
   "claim": "Thinking budget 256 (DT256): quarter budget → −0.028 vs 1024 (CHEAPER-LOSS)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-early-L328"
   ]
  },
  {
   "id": "F-659",
   "kind": "null-result",
   "claim": "Best-of-2 / verifier (DEEPBO2): 35B self-select, 4B verifier → chance-level agreement (no selector)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-early-L331"
   ]
  },
  {
   "id": "F-660",
   "kind": "null-result",
   "claim": "Deep context 1200 (DEEPCTX): larger context → −0.006 (cap barely binds)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-early-L332"
   ]
  },
  {
   "id": "F-661",
   "kind": "null-result",
   "claim": "Vendor sampling presets (DEEPSAMP): temp 1.0 + presence, temp 0.6 → −0.029 / −0.017 (engine defaults stay)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-early-L333"
   ]
  },
  {
   "id": "F-662",
   "kind": "null-result",
   "claim": "DEEPRERANK: bge reranker over the engine's top-20 on the deep path → gold enters top-6 on 2 items (gate 5) (coverage, not order, is the limit (only 12/39 gold items have gold in the top-20))",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-late-L257"
   ]
  },
  {
   "id": "F-663",
   "kind": "null-result",
   "claim": "DEEPITER: planner-driven second retrieval on the deep tier → Opus −0.019 (−0.046…+0.008) (the 35B does not convert more sources into score)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-late-L259"
   ]
  },
  {
   "id": "F-664",
   "kind": "null-result",
   "claim": "DEEPCOVER: prompt the deep model to enumerate key points, longer cap → Opus −0.027; correctness −0.11 (enumerating makes the 35B assert facts it does not hold)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-late-L264"
   ]
  },
  {
   "id": "F-665",
   "kind": "null-result",
   "claim": "QUANTCEIL: 3-bit instead of 2-bit quant of the 35B → V-para two seeds +0.004 (not detected at MDE 0.03; decode ×0.7)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-late-L265"
   ]
  },
  {
   "id": "F-666",
   "kind": "null-result",
   "claim": "DEEPHEDGE: prompt against hedging on the deep tier → fires 0/11 (the \"hedged\" bucket was a canned health fallback, not model hedging)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-late-L268"
   ]
  },
  {
   "id": "F-667",
   "kind": "null-result",
   "claim": "LITE-THINK: thinking on for the 1.7B → 512: −0.041; 256: −0.001 (thinking stays off)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-chron-late-L300"
   ]
  },
  {
   "id": "F-668",
   "kind": "null-result",
   "claim": "R14 Retrieval hurts the 35B?: 0-call",
   "value": "−0.031; coverage miss carries 0.4 % of loss",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:33",
   "caveat": "why closed: mechanism falsified",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-results-L241"
   ]
  },
  {
   "id": "F-669",
   "kind": "null-result",
   "claim": "R17 MoE Qwen3.5-35B closed-book: vs deep35",
   "value": "+0.007 (−0.028…+0.044)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:36",
   "caveat": "why closed: not detected",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-results-L244"
   ]
  },
  {
   "id": "F-670",
   "kind": "null-result",
   "claim": "R18 MoE Gemma 4 26B closed-book: vs deep35",
   "value": "+0.018 (−0.025…+0.057)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:37",
   "caveat": "why closed: not detected",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-results-L245"
   ]
  },
  {
   "id": "F-671",
   "kind": "null-result",
   "claim": "R19 Gemma + thinking vs Qwen + thinking: 2 seeds Opus",
   "value": "−0.028 (−0.057…+0.003)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:38",
   "caveat": "why closed: parity-faster failed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-results-L246"
   ]
  },
  {
   "id": "F-672",
   "kind": "null-result",
   "claim": "MoE 25–35B swap: Qwen3.5-35B, Gemma 26B",
   "value": "+0.007 / +0.018 vs deep35",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md#R17-R18",
   "caveat": "why closed: one plateau ≈ 0.60",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-system-L207"
   ]
  },
  {
   "id": "F-673",
   "kind": "null-result",
   "claim": "Higher deep-tier quant: Q2 → Q3",
   "value": "+0.004 (−0.024…+0.033)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md#R37",
   "caveat": "why closed: gain ≥ +0.03 excluded",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-system-L209"
   ]
  },
  {
   "id": "F-674",
   "kind": "null-result",
   "claim": "Thinking budget 256 / 3072: budget curve",
   "value": "256 −0.028; 3072 +0.010 n.s.",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md#R21-R22",
   "caveat": "why closed: 512 is the knee",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-system-L210"
   ]
  },
  {
   "id": "F-675",
   "kind": "null-result",
   "claim": "Best-of-2 / verifier: 35B self-select, 4B verifier",
   "value": "coin flip (49 % oracle agreement)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md#R27",
   "caveat": "why closed: no selection signal",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-system-L212"
   ]
  },
  {
   "id": "F-676",
   "kind": "null-result",
   "claim": "Lite thinking / Q8 quant: thinking on 1.7B; Q8_0",
   "value": "budget 512 −0.041; Q8 −0.003, decode 0.68×",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: no gain",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-system-L223"
   ]
  },
  {
   "id": "F-677",
   "kind": "null-result",
   "claim": "Upstream LFM2.5-8B-A1B: MoE reader in upstream app",
   "value": "4/17 answers ran out of budget while reasoning",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: rejected by maintainer too",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "N-system-L225"
   ]
  },
  {
   "id": "F-678",
   "kind": "decision",
   "claim": "Build one evaluation harness (ADR-001): own thin protocol + existing scorers; 5 layers; 7 axes never averaged — why: each experiment wrote its own scripts; desktop and device rows had different shapes; no regression gate; run noise ≈ effect size — outcome: harness became boar-eval, later the public boar-bench",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-chron-early-02"
   ]
  },
  {
   "id": "F-679",
   "kind": "decision",
   "claim": "Make the memory-fit estimate repack-aware and load with no_extra_bufts on Android only (why: repack copied ~2.3 GiB of weights into anonymous memory and the phone killed the 4B)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "outcome: Android-only variant frozen; iOS unchanged pending device evidence",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-system-01"
   ]
  },
  {
   "id": "F-680",
   "kind": "event",
   "claim": "iPhone 13 run matches the desktop proxy (0.394 vs 0.395)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 afternoon",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-early-L53"
   ]
  },
  {
   "id": "F-681",
   "kind": "event",
   "claim": "35B memory check in a 6,750 MiB (≈ 6.6 GiB) Linux container: 10/10, 0 OOM",
   "value": "container capped at 6,750 MiB (about 6.6 GiB); memory.peak 6.59 GiB; 10/10 prompts, 0 OOM",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "Verification note: \"6,750 MiB (≈ 6.6 GiB)\" is wrong: the cap was 6,750 MiB. The run peaked essentially at the cap.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-early-L60"
   ]
  },
  {
   "id": "F-682",
   "kind": "fact",
   "claim": "Bounty hard limits: Android/GrapheneOS, ≤ 12 GB RAM, ≤ 50 GB assets, fully offline, no Play Services, public repo",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "n/a",
   "set_status": null,
   "judge": "n/a",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "requirements text, not BOAR's",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-early-02"
   ]
  },
  {
   "id": "F-683",
   "kind": "fact",
   "claim": "Gemma 4 E4B as reader was not adopted: Δ −0.016 inside noise; slow first visible text came from hidden thinking (~18 s), not memory thrash",
   "value": "Δ −0.016",
   "ci": "MDE 0.036, 2×2 seeds",
   "n": null,
   "seeds": "2",
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "harness first counted an empty token as TTFT (fixed)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-early-15"
   ]
  },
  {
   "id": "F-684",
   "kind": "fact",
   "claim": "The desktop proxy matched a real iPhone 13 run",
   "value": "0.394 vs 0.395; Δ −0.001",
   "ci": "(0.353–0.441), MDE 0.038, 2 reps",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "a second Claude-family judge agreed (same winner 99 %, κ 0.894) QUEUE:249",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-early-26"
   ]
  },
  {
   "id": "F-685",
   "kind": "fact",
   "claim": "35B deep tier ran 10/10 in a 6,750 MiB (≈ 6.6 GiB) Linux container with no swap; peak 6.59 GiB",
   "value": "6.59 GiB peak inside a 6,750 MiB (about 6.6 GiB) cap",
   "ci": "10 prompts",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": "n/a",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "macOS \"10 GiB\" was footprint accounting | Verification note: Unit fix: cap was 6,750 MiB, not 6,750 MiB (≈ 6.6 GiB). macOS \"10 GiB\" was footprint accounting.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-early-30"
   ]
  },
  {
   "id": "F-686",
   "kind": "fact",
   "claim": "On an 8 GB Android the 4B was killed by the OS while loading; ARM weight repack copied ~2.3 GiB into anonymous memory",
   "value": "RSS 4.7–5.0 GiB at kill",
   "ci": "2 reproductions",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "debug build; fix tested on desktop only (anon 3.25 → 0.88 GB)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-early-36"
   ]
  },
  {
   "id": "F-687",
   "kind": "fact",
   "claim": "Desktop overstated the 8 GB phone by ~0.03 on the 1.5B",
   "value": "Δ −0.029",
   "ci": "(−0.056…−0.002), MDE 0.036",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "runtime difference suspected",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-early-37"
   ]
  },
  {
   "id": "F-688",
   "kind": "fact",
   "claim": "Same ladder, floor-adjusted: Opus no-web 1.036, Sonnet 0.937, Haiku 4.5 0.564; 4B app 0.387; iPhone 1.5B 0.240.",
   "value": "floor-adj",
   "ci": "see ladder",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-method-24"
   ]
  },
  {
   "id": "F-689",
   "kind": "fact",
   "claim": "Desktop proxy = phone for quality",
   "value": "Galaxy A15 0.561 vs desktop 0.558/0.560 (cut 6); POCO X6 Pro 0.529 = desktop 0.529 (cut 7); iPhone 13 0.394 vs 0.395 (1.5B)",
   "ci": "n 57 each",
   "n": "57",
   "seeds": null,
   "set": "v0.1 (dev)",
   "set_status": "dev",
   "judge": "Opus / Haiku",
   "date": "2026-10-02",
   "date_text": "2026-10-02 → 10-14",
   "source_public": "boar-research:arms/device-x6",
   "caveat": "justifies running research on Macs",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-process-68"
   ]
  },
  {
   "id": "F-690",
   "kind": "fact",
   "claim": "Disk discipline",
   "value": "alarm below 25 GiB free; deletions logged with restore recipe",
   "ci": "~120 logged rows",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 →",
   "source_public": null,
   "caveat": "row count approximate (127 table lines)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-process-70"
   ]
  },
  {
   "id": "F-691",
   "kind": "fact",
   "claim": "The 4B weight repack on ARM duplicates ~2.3 GiB in anonymous memory.",
   "value": "Probe A: anon 3.25 → 0.88–0.91 GB, max RSS 5.1–5.4 → 3.41 GB with no-repack",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "macOS desktop probe",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-57"
   ]
  },
  {
   "id": "F-692",
   "kind": "fact",
   "claim": "iPhone 13 bench numbers for the 1.5B (llama.rn).",
   "value": "pp512 165.8 / tg64 18.3 tok/s; footprint 346–377 MB",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "iOS side line",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-89"
   ]
  },
  {
   "id": "F-693",
   "kind": "event",
   "claim": "Bench v0.2 started; iPhone device run",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-results-L24"
   ]
  },
  {
   "id": "F-694",
   "kind": "event",
   "claim": "Root cause of a 4B kill on an 8 GB moto phone: llama.cpp ARM weight repack copies ~2.3 GiB into anonymous memory",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L26"
   ]
  },
  {
   "id": "F-695",
   "kind": "event",
   "claim": "Probe A: no-repack drops 4B max RSS 5.1–5.4 → 3.41 GB on desktop",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L27"
   ]
  },
  {
   "id": "F-696",
   "kind": "event",
   "claim": "iPhone 13 repack probe: llama.rn does not repack on iOS; fit-repack stays Android-only",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L28"
   ]
  },
  {
   "id": "F-697",
   "kind": "decision",
   "claim": "Strategy memo: Android-only, three fronts (knowledge into the 4B's context; device truth + reader; held-out bench + claim); stop iPhone/1.5B quality work, persona/prompt work, same-class reader swaps — why: measured gap; only the 4B near the bar — outcome: iPhone work stopped; knowledge front closed later that day",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 06:10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-chron-early-06"
   ]
  },
  {
   "id": "F-698",
   "kind": "event",
   "claim": "Strategy memo (high-tier model, one call): Android-only, three fronts; stop iPhone/1.5B and prompt work",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 06:10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-early-L48"
   ]
  },
  {
   "id": "F-699",
   "kind": "event",
   "claim": "Own 8 GB Android: 4B killed by the OS while loading; root cause = ARM weight repack",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 ~18:30",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-early-L61"
   ]
  },
  {
   "id": "F-700",
   "kind": "event",
   "claim": "Deep tier D1 (on-device MoE engine build) done",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-early-L76"
   ]
  },
  {
   "id": "F-701",
   "kind": "fact",
   "claim": "Cut lineage.",
   "value": "cut 1 → 1b → 2 → 3 (VG v3 + AR + snakebite) → 4 (KVFIX + INSTANT-FRAG + INSTANT-SUBJ) → 5 (KV no-source prefix + gate skip) → 6 (DETMAX) → 7 (FACTS r6)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "date_text": "2026-10-03..07",
   "source_public": "boar-research:README.md",
   "caveat": "base = public upstream commit 0cfd4a77",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-16"
   ]
  },
  {
   "id": "F-702",
   "kind": "event",
   "claim": "Deep tier D1: pinned build of the bmoe streamed-MoE engine (Android payload 12.4 MB); deep work paused for token budget",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L30"
   ]
  },
  {
   "id": "F-703",
   "kind": "event",
   "claim": "Ship cut 1 / 1b built (health cards, gate cold-load, travel calc, CO hazard, device detector)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L31"
   ]
  },
  {
   "id": "F-704",
   "kind": "decision",
   "claim": "Focus on the 4B default tier; deep tier parked at D1; coordinator runs work with subagents instead of a standing chat team — why: token budget; no 12 GB phone — outcome: 4B levers exhausted by 10-05; deep tier reopened 10-06",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-chron-early-20"
   ]
  },
  {
   "id": "F-705",
   "kind": "decision",
   "claim": "Fix the KV session save path (KVFIX) and block fragment/figurative instant answers (INSTANT-FRAG) → cut 4 (why: A15 smoke: every question re-prefilled the prefixes; \"capital of France\" answered with a Lyon fragment)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": "outcome: cut 4 built; KVFIX real gate is the device",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-system-05"
   ]
  },
  {
   "id": "F-706",
   "kind": "decision",
   "claim": "Bench APK with INTERNET permission removed (no-network by construction) (why: prove offline operation statically)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": "outcome: audit verdict NO-NETWORK-BY-CONSTRUCTION; later phone runs used it",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-system-06"
   ]
  },
  {
   "id": "F-707",
   "kind": "event",
   "claim": "First real-phone run of a frozen build on a volunteer tester's 8 GB phone (smoke, 9 items)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-early-L84"
   ]
  },
  {
   "id": "F-708",
   "kind": "event",
   "claim": "Cut 4 (KVFIX + instant-answer guards)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-early-L87"
   ]
  },
  {
   "id": "F-709",
   "kind": "event",
   "claim": "No-network measurement APK (INTERNET permission removed by construction)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-early-L89"
   ]
  },
  {
   "id": "F-710",
   "kind": "fact",
   "claim": "Most phone latency came from a KV-session save bug in OUR patch (desktop shim no-op hid it); ~55–65 s re-prefill per answer",
   "value": "55–65 s",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": "not an upstream bug",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-early-61"
   ]
  },
  {
   "id": "F-711",
   "kind": "fact",
   "claim": "Static audit of the first measurement APKs: network-capable (INTERNET declared); a later no-network APK removed every network permission by construction",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": "n/a",
   "date": "2026-10-05",
   "source_public": null,
   "caveat": "measurement app, not the submission app",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-early-66"
   ]
  },
  {
   "id": "F-712",
   "kind": "fact",
   "claim": "Phone latency budget (8 GB phone, warm): decode 42 %, answer prefill 35 %, retrieval 17 %, gate 6 %; p50 ≈ 69 s",
   "value": "68.7 s p50 / 88.9 s p90",
   "ci": "calibrated on 9 rows ±10 %",
   "n": "9",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "n/a",
   "date": "2026-10-05",
   "source_public": null,
   "caveat": "projection for cut 4",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-early-67"
   ]
  },
  {
   "id": "F-713",
   "kind": "fact",
   "claim": "Speculative decoding with a 0.6B draft: acceptance 31 %, phone gain erased by the draft's prefill; llama.rn rc.4 supports only MTP drafts",
   "value": "31.1 %; −8.4 % projected decode",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": "n/a",
   "date": "2026-10-05",
   "source_public": null,
   "caveat": "closed for the 4B",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-early-68"
   ]
  },
  {
   "id": "F-714",
   "kind": "fact",
   "claim": "A15 smoke (cut 3): first phone run matched desktop routing.",
   "value": "9/9 success, tiers 8/8 = desktop, fast TTFT 52–100 s, total up to 147 s, PSS 3.5 GiB with 1.6 GiB swapped",
   "ci": "n 9",
   "n": "9",
   "seeds": null,
   "set": "v0.1 public",
   "set_status": "public dev",
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": "8 GB phone (below 12 GB class)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-26"
   ]
  },
  {
   "id": "F-715",
   "kind": "fact",
   "claim": "KV-cache save bug cost on the A15.",
   "value": "gate prefix 35–41 s + answer prefix 20–28 s re-prefilled per question at ~12 tok/s; fix est. −55 to −65 s",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": "our own regression, not upstream (QUEUE:649)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-27"
   ]
  },
  {
   "id": "F-716",
   "kind": "fact",
   "claim": "q8_0 KV cache is non-inferior on quality.",
   "value": "Δ −0.0077 (−0.0232…+0.0062); reader KV 576 → 306 MiB",
   "ci": "n 60 × 2 seeds",
   "n": "60",
   "seeds": "2",
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-05",
   "source_public": null,
   "caveat": "desktop slower; device-day candidate only | device candidate",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-58",
    "F-chron-early-65"
   ]
  },
  {
   "id": "F-717",
   "kind": "event",
   "claim": "Cut 4 (KV-cache fix + instant fragment) built; cut 5 (lite gate skip) adopted as device candidate",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "date_text": "2026-10-05 (cut 4) / 2026-10-06 (cut 5)",
   "source_public": null,
   "caveat": "cut 5 (CUT5-LITE) was adopted on 2026-10-06, not 10-05",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-results-L33"
   ]
  },
  {
   "id": "F-718",
   "kind": "event",
   "claim": "First real-phone run (cut 3 smoke, Galaxy A15 8 GB): 9/9, tree hash = desktop, latency 52–147 s",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L36"
   ]
  },
  {
   "id": "F-719",
   "kind": "event",
   "claim": "A15 diagnosis: KV-cache save path bug (our own regression) re-prefills ~430+~300 tokens per question",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L37"
   ]
  },
  {
   "id": "F-720",
   "kind": "event",
   "claim": "Cut 4 = cut 3 + KVFIX + INSTANT-FRAG (+ INSTANT-SUBJ)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L38"
   ]
  },
  {
   "id": "F-721",
   "kind": "event",
   "claim": "No-network APK by construction (INTERNET permission removed)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L39"
   ]
  },
  {
   "id": "F-722",
   "kind": "decision",
   "claim": "Adopt CUT5-LITE (KV prefix for no-source prompts + route-gate skip), reject CUT5-SPEED (why: CUT5-SPEED failed its desktop speed gate (1.071× vs ≤ 0.8); lite was identity-safe)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "outcome: cut 5 = device-day candidate",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-system-07"
   ]
  },
  {
   "id": "F-723",
   "kind": "decision",
   "claim": "Freeze claim/device build at cut 6; only routing/safety fixes later with v0.1 answers unchanged (why: protect held-out numbers)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "superseded on 10-07: cut 7 (FACTS r6) changed 8 v0.1 answers, against the 'v0.1 answers unchanged' clause (CUT7.md:29; QUEUE l.770)",
   "verification": "CONFLICT",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-system-09"
   ]
  },
  {
   "id": "F-724",
   "kind": "event",
   "claim": "Phone (8 GB) full run, cut 6: 73/73, phone 0.561 = desktop 0.558/0.560",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-early-L91"
   ]
  },
  {
   "id": "F-725",
   "kind": "fact",
   "claim": "Partial phone quality (cut 3) on 25 v0.1 items equalled desktop",
   "value": "0.482 vs 0.478 / 0.493",
   "ci": "(0.402–0.563), 25 items",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus-med",
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "later found invalid (wrong refs) — superseded by F-chron-early-70",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-early-69"
   ]
  },
  {
   "id": "F-726",
   "kind": "fact",
   "claim": "A15 latency budget for cut 4 (model).",
   "value": "warm p50 68.7 s / p90 88.9 s; cold 116.8 / 141.7 s; warm shares: decode 42 %, answer prefill 35 %, retrieval 17 %, gate 6 %",
   "ci": "±10 % calibration on 9 rows",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "model, not a measurement",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-28"
   ]
  },
  {
   "id": "F-727",
   "kind": "fact",
   "claim": "A15 device rates.",
   "value": "prefill 13–14 tok/s, decode 2.0–2.7 tok/s; flash attention off: decode 3.7–3.8 but prefill −25 %",
   "ci": "9 runs",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "thread ranking within noise",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-29"
   ]
  },
  {
   "id": "F-728",
   "kind": "fact",
   "claim": "Route-gate skip (cut 5) skips 32/54 gate calls per seed with identical answers.",
   "value": "59 % skipped; 120/120 identical text/tier/passages",
   "ci": "2 seeds",
   "n": null,
   "seeds": "2",
   "set": "v0.1",
   "set_status": null,
   "judge": "none",
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "phone gain only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-59"
   ]
  },
  {
   "id": "F-729",
   "kind": "fact",
   "claim": "Retrieval perf arms show no desktop gain but large phone gain (confounded).",
   "value": "desktop 1.071×; A15 retrieval 60.8 → 27.1 s (0.445×)",
   "ci": "n 1 each",
   "n": "1",
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "order/cache confound, needs ABBA",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-60"
   ]
  },
  {
   "id": "F-730",
   "kind": "event",
   "claim": "A15 latency budget: warm p50 68.7 s model-path answer predicted for cut 4",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L41"
   ]
  },
  {
   "id": "F-731",
   "kind": "event",
   "claim": "Claim/device build frozen at cut 6; later only routing/safety fixes",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "the 10-06 freeze ('later only routing/safety fixes with v0.1 answers unchanged', QUEUE l.701) was relaxed on 10-07 when cut 7 added FACTS r6, which changes 8 v0.1 answers (CUT7.md:29; QUEUE l.770). Publish both steps",
   "verification": "CONFLICT",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L44"
   ]
  },
  {
   "id": "F-732",
   "kind": "event",
   "claim": "Cut 6 on the A15, full run: 73/73, phone 0.561 vs desktop 0.558/0.560",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L46"
   ]
  },
  {
   "id": "F-733",
   "kind": "decision",
   "claim": "\"No device exists\": device tasks dormant — why: No 12 GB phone available — outcome: Later reversed in practice when a 10.7 GiB phone ran cut 7 on 10-14",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-process-10"
   ]
  },
  {
   "id": "F-734",
   "kind": "event",
   "claim": "Venue path proven with device location + area tile: 0/7 → 7/7",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-early-L102"
   ]
  },
  {
   "id": "F-735",
   "kind": "event",
   "claim": "On-device MoE engine judged-equivalent to the desktop server (DEEPPAR-L1)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-early-L104"
   ]
  },
  {
   "id": "F-736",
   "kind": "event",
   "claim": "Haiku 5.5-low becomes the L1 triage judge; judge sharding N = 4; \"no device exists\" ruling",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-process-L40"
   ]
  },
  {
   "id": "F-737",
   "kind": "fact",
   "claim": "Venue questions: 0/7 → 7/7 with device location + the user's area map tile (+3.65 MB), no engine change",
   "value": "7/7 both seeds",
   "ci": "7 items × 2 seeds",
   "n": null,
   "seeds": "2",
   "set": "V-para venue items (dev-style)",
   "set_status": "dev-style",
   "judge": "OSM gold match",
   "date": "2026-10-07",
   "source_public": null,
   "caveat": "gold and engine both from OSM: proves the path, not venue quality",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-early-85"
   ]
  },
  {
   "id": "F-738",
   "kind": "fact",
   "claim": "Venue questions with device location + area tile",
   "value": "0/7 → 7/7 both seeds",
   "ci": "7 items",
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para (burnt for this)",
   "set_status": "dev-style (V-para after 2026-10-06)",
   "judge": "deterministic scorer",
   "date": "2026-10-07",
   "source_public": null,
   "caveat": "'burnt for this' is not the ruling; gold and venues both OSM: plumbing proof only",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-results-104"
   ]
  },
  {
   "id": "F-739",
   "kind": "fact",
   "claim": "FACTS speed.",
   "value": "warm p95 0.090 ms; cold load 111–129 ms (desktop)",
   "ci": "n 5,500 warm",
   "n": "5,500",
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": "phone cold load unmeasured",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-22"
   ]
  },
  {
   "id": "F-740",
   "kind": "fact",
   "claim": "bmoe desktop speed and memory.",
   "value": "prefill 25.7 s median, decode 8.6 tok/s, session peak 11.7 GiB on macOS (Linux anon 4.0–4.4 GiB)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": "accounting differs by OS",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-47"
   ]
  },
  {
   "id": "F-741",
   "kind": "fact",
   "claim": "Venue answers need device location plus the user's area tile.",
   "value": "P5 0/7 → 7/7 both seeds; +3.65 MB tiles; instant 5–192 ms",
   "ci": "n 7, 2 seeds",
   "n": "7,",
   "seeds": "2",
   "set": "V-para venue items",
   "set_status": null,
   "judge": "none",
   "date": "2026-10-07",
   "source_public": null,
   "caveat": "venues and gold both OSM tags: proves the path, not venue quality",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-76"
   ]
  },
  {
   "id": "F-742",
   "kind": "event",
   "claim": "VENUEGPS-2: venue items 0/7 → 7/7 with device location + area tile (plumbing proof)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-results-L42"
   ]
  },
  {
   "id": "F-743",
   "kind": "event",
   "claim": "DEEPPAR / DEEPPAR-L1: bmoe judged-equivalent to llama-server",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L48"
   ]
  },
  {
   "id": "F-744",
   "kind": "event",
   "claim": "VENUEGPS-2: venue items 0/7 → 7/7 with device location + area tile",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L49"
   ]
  },
  {
   "id": "F-745",
   "kind": "decision",
   "claim": "Cut 7 = cut 6 + FACTS r6 (adopted on gates: 0 fires on held-out sets); no relaxed lowercase rule; fine-tune data 1k go; \"no device exists\" — why: deterministic facts lever; no phone — outcome: cut 7 frozen; fine-tune data failed review same day",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "date_text": "2026-10-07 09:23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-chron-early-26"
   ]
  },
  {
   "id": "F-746",
   "kind": "event",
   "claim": "Cut 7 = cut 6 + FACTS r6; v0.3 to be written outside the Claude family; \"no device exists\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "date_text": "2026-10-07 09:23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-early-L99"
   ]
  },
  {
   "id": "F-747",
   "kind": "event",
   "claim": "DEEPPAR-L1: on-device MoE engine (bmoe) judged-equivalent to llama-server (Δ −0.012) → TRANSFER",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "date_text": "2026-10-07 22:50",
   "source_public": "boar-research:reports/CLOSED.md:35",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-results-L43"
   ]
  },
  {
   "id": "F-748",
   "kind": "decision",
   "claim": "Recommended deep config = cut 6 + Qwen3.6-35B-A3B UD-Q2_K_XL + thinking budget 512 (why: 512 = 1024 in quality at 0.55× time)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "source_public": null,
   "caveat": "outcome: not in claim build; needs a 12 GB device and an app-side thinking cap",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-system-11"
   ]
  },
  {
   "id": "F-749",
   "kind": "fact",
   "claim": "Phone time for the deep tier from third-party public rates (4–6 tok/s decode): 3.5–5 min per answer at thinking budget 1024 (2026-10-09 estimate); corrected on 2026-10-12 to ≈ 1.9–2.8 min at budget 512.",
   "value": "On 10-09 the estimate was 3.5-5 min per answer at thinking budget 1024 (4-6 tok/s decode); the 1.9-2.8 min figure at budget 512 is the 2026-10-12 deep-research correction",
   "ci": "external",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": "n/a",
   "date": "2026-10-09",
   "source_public": null,
   "caveat": "competitor's numbers, our arithmetic | Verification note: Anachronism: the 512-budget phone time did not exist on 10-09. Competitor phone rates + our arithmetic, never measured by BOAR.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-early-99"
   ]
  },
  {
   "id": "F-750",
   "kind": "fact",
   "claim": "Gemma 4 26B-A4B + thinking (G) on V-para",
   "value": "raw 0.514; floor-adj 0.388",
   "ci": "raw 0.487–0.540; floor 0.354–0.421; 2 seeds",
   "n": "51 (inferred from set)",
   "seeds": "2",
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-09",
   "source_public": "boar-research:reports/BAR-TABLE.md:42",
   "caveat": "27 % faster than T, RSS ≈ 7.7 GB",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-results-32"
   ]
  },
  {
   "id": "F-751",
   "kind": "fact",
   "claim": "bmoe v0.28.0 has no thinking-budget flag; a device needs an app-side cap.",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "source_public": null,
   "caveat": "unbounded thinking ≈ 1.3k tokens, ≈ 2× time",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-49"
   ]
  },
  {
   "id": "F-752",
   "kind": "decision",
   "claim": "Fail-closed pipeline: errored traces/missing ids stop the run or score at the floor (why: the old judge silently dropped errored traces, flattering a failing device)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": "boar-research:reports/RULE-CHANGES.md:19; boar-bench:harness/judge.mjs:8-10",
   "caveat": "outcome: flags `--strict`, `--floor-errored`, `--floor-missing`",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-method-20"
   ]
  },
  {
   "id": "F-753",
   "kind": "decision",
   "claim": "Win-rate reported, never gated (later: proposal to drop the win-rate conjunct because it is saturated at 0) (why: every phone system scores 0 wins vs a 4.9/5 reference)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 (BAR-TABLE); plan 2026-10-02 lists W/T/L as a secondary metric",
   "source_public": "boar-bench:bench/v0.3/plan/analysis-plan.md:30; boar-research:reports/BAR-TABLE.md:7",
   "caveat": "the plan never gated on win-rate, but the deep-tier arm reports of 10-08/09 read the bar as 'floor-adj CI low > 0.50 AND win-rate > 0'; 'reported, never gated' was fixed in BAR-TABLE on 10-10 after RND-REVIEW B1. State both",
   "verification": "CONFLICT",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-method-28"
   ]
  },
  {
   "id": "F-754",
   "kind": "fact",
   "claim": "Cut 7 no-network measurement APK built: engine tree identical to desktop cut 7, 0 network permissions",
   "value": "tree 7c96e59e…",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": "n/a",
   "date": "2026-10-10",
   "source_public": null,
   "caveat": "measurement app, not the chat app",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-early-108"
   ]
  },
  {
   "id": "F-755",
   "kind": "fact",
   "claim": "Win-rate vs the reference under the Claude judge is 0 on every row and seed of V-para and v0.2; isolated wins exist only on the public dev set v0.1.",
   "value": "0 wins on every row and seed of V-para and v0.2 under Opus; isolated wins exist on v0.1 dev (1/57 under Opus for the deep35 R450 and 4B rows; 2/57 under Haiku on the phone run)",
   "ci": "all rows",
   "n": null,
   "seeds": null,
   "set": "all",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:80; boar-research:arms/device-x6/DEVICE-X6.md:27-28",
   "caveat": "'every set' is wrong: scope the zero to V-para and v0.2. The public TRACE-FORMAT.md:70 line '0 wins on every set so far' has the same scope problem",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-method-57"
   ]
  },
  {
   "id": "F-756",
   "kind": "fact",
   "claim": "The same 35B model file was deleted twice under queued experiments (once by an app-packaging disk cleanup, once by a model swap), costing two 12.3 GB re-downloads.",
   "value": "2 × 12.3 GB re-downloads",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10, 2026-10-12",
   "source_public": "boar-research:reports/RULE-CHANGES.md:29",
   "caveat": "→ rule 22 | Verification note: The same 35B Q2_K_XL file was deleted twice: 10-10 22:42 by the APK track's disk cleanup, 10-12 08:21 by the coordinator's QUANTCEIL model swap (not housekeeping both times).",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-process-43"
   ]
  },
  {
   "id": "F-757",
   "kind": "event",
   "claim": "Deep tier handoff note: config + engine changes needed; no 12 GB device",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": "the note is marked 'DRAFT: send only after [the author] approves' (DEEP-TIER-2026-10-10.md:2): call it a draft hand-off, not a release",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L51"
   ]
  },
  {
   "id": "F-758",
   "kind": "decision",
   "claim": "Proceed on stated defaults for the claim path (debug-signed APK, raw formal + floor-adj + win-rate, no v0.3, no device relay) — why: unblock work without waiting on the author — outcome: superseded 45 min later by the STOP",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 21:05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-chron-late-01"
   ]
  },
  {
   "id": "F-759",
   "kind": "event",
   "claim": "Decision note with defaults sent to the author (claim vehicle, bar scale \"raw formal + floor-adj + win-rate\", no v0.3, no device relay)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 21:05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-late-L19"
   ]
  },
  {
   "id": "F-760",
   "kind": "event",
   "claim": "Cut 7 no-network measurement APK built and audited",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 21:27",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-early-L115"
   ]
  },
  {
   "id": "F-761",
   "kind": "event",
   "claim": "Measurement-app (\"BOAR Bench\") cut 7 no-network APK built and audited: no INTERNET permission",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 21:27",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-late-L21"
   ]
  },
  {
   "id": "F-762",
   "kind": "event",
   "claim": "Android chat wired to the measured engine on a branch; identity check 20/20",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 21:40",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-late-L22"
   ]
  },
  {
   "id": "F-763",
   "kind": "decision",
   "claim": "Research on the Mac mini only until the bar is met; no APK/device/claim-package work — why: author's direction after the process review — outcome: handoff rewritten for research scope",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 21:54",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-chron-early-34"
   ]
  },
  {
   "id": "F-764",
   "kind": "event",
   "claim": "Author's direction: research on the Mac mini only until the bar is met; device and claim work after",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 21:54",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-late-L24"
   ]
  },
  {
   "id": "F-765",
   "kind": "fact",
   "claim": "Peer deep research corrected the phone time of the deep config",
   "value": "≈ 1.9–2.8 min per answer on a Pixel 8 Pro CPU at budget 512",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": "boar-research:reports/DEEP-RESEARCH-2026-10-12.md",
   "caveat": "derived from a competitor's public phone numbers, not our measurement",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-late-91"
   ]
  },
  {
   "id": "F-766",
   "kind": "fact",
   "claim": "Deep + think 512 on a Pixel 8 Pro CPU is projected at 1.9–2.8 min per answer",
   "value": "1.9–2.8 min",
   "ci": "from published third-party rates",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": "boar-research:reports/DEEP-RESEARCH-2026-10-12.md:8",
   "caveat": "not our measurement; deep tier never ran on a phone",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-results-68"
   ]
  },
  {
   "id": "F-767",
   "kind": "fact",
   "claim": "Phone time estimate for deep + think 512.",
   "value": "≈ 1.9–2.8 min per answer (Pixel 8 Pro CPU, from third-party rates 4–6 tok/s decode, 32–36 prefill)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": "boar-research:reports/DEEP-RESEARCH-2026-10-12.md:8,24",
   "caveat": "estimate, not a measurement",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-50"
   ]
  },
  {
   "id": "F-768",
   "kind": "decision",
   "claim": "Phone-time wording: deep + think 512 ≈ 1.9–2.8 min on a Pixel 8 Pro CPU; competitor percentages never quoted as a gap — why: corrected premises 2 and 5 — outcome: applied in all later texts",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 14:14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-chron-late-09"
   ]
  },
  {
   "id": "F-769",
   "kind": "event",
   "claim": "Phone tester package for a 12 GB-class phone built and shared via a link folder",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 21:14 / 21:4x",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-late-L59"
   ]
  },
  {
   "id": "F-771",
   "kind": "fact",
   "claim": "Phone speed (cut 7, POCO X6 Pro, MT6897, 10.7 GiB)",
   "value": "fast tier p50 30.7 s / p90 48.4 s; TTFT p50 18.5 s; decode 5.5 tok/s; all 72 p50 24.7 s",
   "ci": "72/72 answered",
   "n": null,
   "seeds": null,
   "set": "v0.1 public",
   "set_status": "public dev",
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:arms/device-x6/DEVICE-X6.md:15-17",
   "caveat": "airplane mode, no INTERNET permission",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-results-70"
   ]
  },
  {
   "id": "F-772",
   "kind": "fact",
   "claim": "Phone memory (cut 7)",
   "value": "PSS peak 5.88 GB at load, then p50 3.64 / max 4.23 GB",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:arms/device-x6/DEVICE-X6.md:18",
   "caveat": "thermal status 0 over 29 min",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-results-71"
   ]
  },
  {
   "id": "F-773",
   "kind": "fact",
   "claim": "Phone vs desktop speed ratio",
   "value": "total p50 28.3 s vs 7.5 s (×3.4); decode 5.6 vs 22.6 tok/s",
   "ci": "60 ids",
   "n": null,
   "seeds": null,
   "set": "v0.1",
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:arms/device-x6/DEVICE-X6.md:22",
   "caveat": "desktop = MacBook M1 Pro",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-results-72"
   ]
  },
  {
   "id": "F-775",
   "kind": "fact",
   "claim": "Cut 7 on the POCO X6 Pro (MT6897, 10.7 GiB, Android 16, airplane, no INTERNET permission).",
   "value": "72/72 success, 0 timeouts, logcat clean",
   "ci": "n 72 + warm-up",
   "n": "72",
   "seeds": null,
   "set": "v0.1 public + safety",
   "set_status": "public dev",
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:arms/device-x6/DEVICE-X6.md:5-13",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-31"
   ]
  },
  {
   "id": "F-776",
   "kind": "fact",
   "claim": "X6 fast-tier latency.",
   "value": "total p50 30.7 s, p90 48.4 s, max 53.0 s; TTFT p50 18.5 s; decode p50 5.5 tok/s (4.6–7.4); output p50 64 tokens",
   "ci": "n 51 fast",
   "n": "51",
   "seeds": null,
   "set": "v0.1",
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:arms/device-x6/DEVICE-X6.md:15",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-32"
   ]
  },
  {
   "id": "F-777",
   "kind": "fact",
   "claim": "X6 instant tier.",
   "value": "p50 0.1 s, p90 11.2 s; 21 instant (guard 12, extractive 4, places 3, calculator 2)",
   "ci": "n 21",
   "n": "21",
   "seeds": null,
   "set": "v0.1",
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:arms/device-x6/DEVICE-X6.md:14,16",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-33"
   ]
  },
  {
   "id": "F-778",
   "kind": "fact",
   "claim": "X6 memory.",
   "value": "PSS peak 5.88 GB during model load, then p50 3.64 / p90 3.80 / max 4.23 GB",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:arms/device-x6/DEVICE-X6.md:18",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-34"
   ]
  },
  {
   "id": "F-779",
   "kind": "fact",
   "claim": "X6 thermal.",
   "value": "battery max 41.7 °C, thermal_status 0 throughout",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:arms/device-x6/DEVICE-X6.md:19",
   "caveat": "charging during run",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-35"
   ]
  },
  {
   "id": "F-780",
   "kind": "fact",
   "claim": "X6 parity with desktop.",
   "value": "tier 60/60, instant 16/16 byte-identical, kept titles 59/60, 4B text identical 3/44 (sampler differences); phone 3.4× slower (decode 22.6 vs 5.6 tok/s)",
   "ci": "n 60",
   "n": "60",
   "seeds": null,
   "set": "v0.1",
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:arms/device-x6/DEVICE-X6.md:22",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-36"
   ]
  },
  {
   "id": "F-781",
   "kind": "fact",
   "claim": "X6 judged quality: device = desktop.",
   "value": "raw 0.529 (0.481–0.575) vs 0.529 (0.484–0.575); floor-adj 0.403 vs 0.402; paired −0.000 (−0.026…+0.028); win/tie/loss 2/1/54 vs 1/2/54",
   "ci": "n 57, 1 run",
   "n": "57,",
   "seeds": null,
   "set": "v0.1 public (dev)",
   "set_status": "public dev",
   "judge": "Haiku L1",
   "date": "2026-10-14",
   "source_public": "boar-research:arms/device-x6/DEVICE-X6.md:26-30",
   "caveat": "Haiku vs A15's Opus: levels not comparable",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-37"
   ]
  },
  {
   "id": "F-783",
   "kind": "event",
   "claim": "DEVICE-X6: cut 7 on POCO X6 Pro, phone 0.529 = desktop 0.529 (Haiku, v0.1 dev), fast-tier p50 31 s, app PSS peak 5.88 GB during load then ≤ 4.23 GB",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:arms/device-x6/DEVICE-X6.md:15-30",
   "caveat": "'PSS ≤ 4.23 GB' must say 'after load'; the load peak is 5.88 GB",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-results-L64"
   ]
  },
  {
   "id": "F-784",
   "kind": "event",
   "claim": "Cut 7 on a 12 GB-class phone (POCO X6 Pro): device = desktop",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:arms/device-x6/DEVICE-X6.md:1-30",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L58"
   ]
  },
  {
   "id": "F-788",
   "kind": "decision",
   "claim": "Official submission vehicle = the maintainer's native Kotlin app (two APKs: ~30 MB onboarding; ~2.7 GB zero-network bundle) (why: maintainer's choice; upstream main becomes an archive)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": "outcome: research engine transfers as port specs; native app benchmarked by the research team",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-system-16"
   ],
   "superseded_by": [
    "F-1703",
    "F-1686"
   ],
   "superseded_note": "The submission changed on 2026-10-24: BoarLite, a new closed app by the author, replaces the app maintainer's native app as the bounty submission."
  },
  {
   "id": "F-789",
   "kind": "fact",
   "claim": "Submission vehicle: maintainer's native Kotlin app, two APKs.",
   "value": "~30 MB onboarding; ~2.7 GB zero-network bundle (Qwen3-4B + bge-small + Vital Articles)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": "maintainer's figures",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-83"
   ]
  },
  {
   "id": "F-790",
   "kind": "decision",
   "claim": "Lite tier is post-10-20 work, explored with quick diagnostic tests; starts with an unsafe-answer warning — why: 2B −0.168 and 6/15 unsafe health answers; weaker phones matter — outcome: LITE2–LITE8, LITEREADERS etc. 10-15/16; closed for now",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 22:21–23:0x",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-chron-late-33"
   ],
   "superseded_by": [
    "F-1386",
    "F-1349"
   ],
   "superseded_note": "Superseded on 2026-10-20: Qwen3-1.7B became the main 8 GB engine (not only a lite tier for 4-6 GB phones); the 4B became the optional tier."
  },
  {
   "id": "F-791",
   "kind": "fact",
   "claim": "Ternary 2B as lite model (BONSAI2-2B)",
   "value": "−0.168 raw vs 4B; floor-adj 0.210 vs 0.420; 6/15 unsafe model-only health answers vs 0 for 4B",
   "ci": "(−0.238…−0.105); n 20",
   "n": "20",
   "seeds": null,
   "set": "v0.2 random20 dev",
   "set_status": "dev",
   "judge": "Haiku L1",
   "date": "2026-10-15",
   "date_text": "2026-10-15 22:36",
   "source_public": null,
   "caveat": "decode 47 vs 25 tok/s",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-late-65"
   ]
  },
  {
   "id": "F-792",
   "kind": "decision",
   "claim": "Expert streaming (LES fork) supersedes earlier loader/kernel candidates on paper for a deep-tier device plan (why: 2.8 GiB RSS, judged parity)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "outcome: phone flash speed unmeasured; Android buffer unbuilt",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "D-system-20"
   ]
  },
  {
   "id": "F-793",
   "kind": "fact",
   "claim": "LES (expert streaming) on a phone is projected at ≈ 4 tok/s (UFS 3.1) to ≈ 6 tok/s (UFS 4.0) including compute; analytic, unmeasured.",
   "value": "with compute ≈ 4 tok/s (UFS 3.1) and ≈ 6 tok/s (UFS 4.0); I/O-only upper bounds 4.8 / 8.0 tok/s",
   "ci": "analytic",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "'6-8' mixes the compute-included estimate with the I/O-only bound; analytic, unmeasured",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-results-67"
   ]
  },
  {
   "id": "F-794",
   "kind": "fact",
   "claim": "LES fork judged parity with llama-server.",
   "value": "0.625 (0.581–0.670) vs 0.634; Δ −0.010 (−0.042…+0.018), floor-adj 0.528 vs 0.541 → TRANSFER",
   "ci": "n 57, 1 seed, temp 0.7",
   "n": "57,",
   "seeds": "1",
   "set": "v0.1 dev, identical contexts",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "noise floor ≈ 0.05",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-53"
   ]
  },
  {
   "id": "F-795",
   "kind": "fact",
   "claim": "Phone flash bound for expert streaming.",
   "value": "320 expert reads/token, 314 MiB/token, 0 cache hits → ≈ 4 tok/s (UFS 3.1) to ≈ 6–8 tok/s (UFS 4.0)",
   "ci": "analytic",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "not measured on a phone",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-54"
   ]
  },
  {
   "id": "F-796",
   "kind": "fact",
   "claim": "Thinking works on the LES fork with the 512 budget forced (llama-cli path only).",
   "value": "10/10 think blocks closed at 512; ≈ 56 s per item wall; RSS 2.9 GiB",
   "ci": "n 10",
   "n": "10",
   "seeds": null,
   "set": "public prompts",
   "set_status": "public dev",
   "judge": "none",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "no quality claim",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-system-55"
   ]
  },
  {
   "id": "F-797",
   "kind": "event",
   "claim": "LESPORT: expert-streaming fork runs the 35B in 2.8 GiB RSS on desktop CPU, judged parity TRANSFER",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-system-L65"
   ]
  },
  {
   "id": "F-798",
   "kind": "event",
   "claim": "LESPORT: expert-streaming fork runs the 35B in 2.8 GiB; judged-equivalent on 57 items",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 18:13–20:11",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "E-chron-late-L93"
   ]
  },
  {
   "id": "F-799",
   "kind": "fact",
   "claim": "Expert streaming (LES fork) runs the 35B with little resident memory",
   "value": "14–17 tok/s at ≈ 2.8 GiB RSS on the mini CPU; lossless vs the fork's own base 10/10",
   "ci": "10 prompts",
   "n": null,
   "seeds": null,
   "set": "public prompts",
   "set_status": null,
   "judge": "none",
   "date": "2026-10-16",
   "date_text": "2026-10-16 18:13 / 19:32",
   "source_public": null,
   "caveat": "first measured with experts in page cache (≈ 12.5 GiB actually in use); phone unmeasured",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-late-73"
   ]
  },
  {
   "id": "F-800",
   "kind": "event",
   "claim": "LESPORT-L1: expert-streaming fork judged-equivalent (−0.010); 2.8 GiB RSS on desktop",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 20:10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "T-results-L73"
   ]
  },
  {
   "id": "F-801",
   "kind": "fact",
   "claim": "Analytic phone speed for streaming",
   "value": "≈ 4–8 tok/s (flash-bound, 314 MiB read per token)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 20:11",
   "source_public": null,
   "caveat": "analytic, not measured",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-late-75"
   ]
  },
  {
   "id": "F-802",
   "kind": "fact",
   "claim": "Hardware used",
   "value": "Mac mini M4 16 GB (long generations, 35B model); MacBook M1 Pro 16 GB (replays, 4B runs, judge shards); phones: Galaxy A15 8 GB, POCO X6 Pro 10.7 GiB, Pixel 8a 8 GB, iPhone 13; no 12 GB phone for most of the project",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "source_public": null,
   "caveat": "device names are fine; no hostnames",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-process-69"
   ]
  },
  {
   "id": "F-807",
   "kind": "fact",
   "claim": "Maintainer's own relevance-floor test on the phone",
   "value": "sources on 38 / 54 / 72 of 73 questions at floor 0.70 / 0.65 / 0.60; median 18.4 / 29.3 / 32.0 s",
   "ci": "counts only, no grader",
   "n": null,
   "seeds": null,
   "set": "v0.1",
   "set_status": null,
   "judge": "none",
   "date": "2026-10-17",
   "date_text": "2026-10-17 14:03",
   "source_public": null,
   "caveat": "his measurement; he kept 0.70",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "F-chron-late-97"
   ]
  },
  {
   "id": "F-808",
   "kind": "null-result",
   "claim": "Speculative decoding (SPECDEC): 0.6B draft → 31 % acceptance; net loss on phone (runtime supports MTP only)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "N-chron-early-L312"
   ]
  },
  {
   "id": "F-809",
   "kind": "null-result",
   "claim": "Desktop speed gate for phone perf arms (CUT5-SPEED): native zstd, SQL batching on desktop → retrieval 1.07× (desktop can't see phone gain) (gate design blind to phone)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "N-chron-early-L313"
   ]
  },
  {
   "id": "F-810",
   "kind": "null-result",
   "claim": "bmoe n-gram / prefetch speed knobs (DEEPSPEED): --ngram, --prefetch → ngram text differs 8/10; slower (no knob adopted)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:55",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "N-chron-early-L324"
   ]
  },
  {
   "id": "F-811",
   "kind": "null-result",
   "claim": "Win-rate as a bar conjunct: \"CI low > 0.50 AND win-rate > 0\"",
   "value": "0 wins for every phone system",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: cannot discriminate; reported, never gated",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "N-method-L218"
   ]
  },
  {
   "id": "F-812",
   "kind": "null-result",
   "claim": "Reproducing claim numbers from the public repo alone: third-party rerun",
   "value": "not possible for v0.2/V-para (held-out, hashes only); possible for v0.1",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-bench:README.md:50",
   "caveat": "cited l.64-65 are about phone time; at 10-13 the red team said NO for all sets (v0.1 references were gitignored too); a v0.1 rerun became possible only with the public bench release that re-issued the references",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "N-method-L226"
   ]
  },
  {
   "id": "F-813",
   "kind": "null-result",
   "claim": "R16 bmoe vs llama-server parity: judged on identical contexts",
   "value": "−0.012 (−0.045…+0.020)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:35",
   "caveat": "why closed: equivalence shown (not a lever)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "N-results-L243"
   ]
  },
  {
   "id": "F-814",
   "kind": "null-result",
   "claim": "R28 Per-item routing 4B ↔ deep: 0 calls",
   "value": "43 %/34 % capture out-of-sample",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:47",
   "caveat": "why closed: chance-level; needs 12 GB device",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "N-results-L255"
   ]
  },
  {
   "id": "F-815",
   "kind": "null-result",
   "claim": "R35 Speculative decoding MTP: external phone numbers",
   "value": "5.6 vs 5.8–6.1 tok/s (slower)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:54",
   "caveat": "why closed: external evidence",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "N-results-L262"
   ]
  },
  {
   "id": "F-816",
   "kind": "null-result",
   "claim": "R36 n-gram speculation in bmoe: 10 prompts",
   "value": "text identical 2/10 (broken)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:55",
   "caveat": "why closed: closed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "N-results-L263"
   ]
  },
  {
   "id": "F-817",
   "kind": "null-result",
   "claim": "R75 Cut 7 on 12 GB phone: measurement",
   "value": "phone = desktop",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:104",
   "caveat": "why closed: not a lever (measured)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "N-results-L296"
   ]
  },
  {
   "id": "F-818",
   "kind": "null-result",
   "claim": "Retrieval perf arms on desktop (CUT5-SPEED): SQL batching, vector cache, native zstd",
   "value": "desktop retrieval 1.071× (bar ≤ 0.8)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: node/SQLite has no bridge/zstd cost to remove; value only on phone",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "N-system-L199"
   ]
  },
  {
   "id": "F-819",
   "kind": "null-result",
   "claim": "Speculative decoding on the 4B: Qwen3-0.6B draft",
   "value": "acceptance 31.1 %, projected A15 decode −8.4 % (bar 20 %), +1.03 GiB, llama.rn rc.4 lacks draft-model path",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: below bar and unsupported",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "N-system-L200"
   ]
  },
  {
   "id": "F-820",
   "kind": "null-result",
   "claim": "q8 KV on desktop: KV cache q8_0",
   "value": "quality equal, but desktop prefill +14 %",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: device decides; not adopted",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "N-system-L203"
   ]
  },
  {
   "id": "F-821",
   "kind": "null-result",
   "claim": "Thread count on A15: t2/t4/t6/t8 sweep",
   "value": "ranking within run-to-run noise (13 % drift)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: no clear winner",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "N-system-L204"
   ]
  },
  {
   "id": "F-822",
   "kind": "null-result",
   "claim": "bmoe speed knobs: --ngram, --prefetch 2, --mtp",
   "value": "ngram not token-identical (2/10), prefetch ×0.94, no MTP block",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: no knob adopted",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "N-system-L208"
   ]
  },
  {
   "id": "F-823",
   "kind": "decision",
   "claim": "Research agents work in fixed lanes, write one \"technique card\" per idea, never run experiments or touch the product; only the coordinator queues runs — why: Avoid duplicate work; separate proposing from measuring — outcome: the retrieval research agent 19 cards, the models/latency research agent 22, the evaluation/safety research agent 23; runner agent executed",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-process-01"
   ]
  },
  {
   "id": "F-824",
   "kind": "event",
   "claim": "Research mandate: three researcher lanes (retrieval, models, evaluation/safety) write \"technique cards\"; a runner executes arms",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-early-L40"
   ]
  },
  {
   "id": "F-825",
   "kind": "fact",
   "claim": "Research was split into named agent lanes: retrieval/knowledge, models/inference, truthfulness/safety/eval, native performance, a runner, and an on-call strategy agent",
   "value": "6 roles",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "source_public": null,
   "caveat": "agent names are internal; describe as \"research agents with lanes\"",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-01"
   ]
  },
  {
   "id": "F-826",
   "kind": "event",
   "claim": "Research lanes start: the models/latency research agent cycle 1 (21:06), the evaluation/safety research agent cycle 1 (21:10), the retrieval research agent cycle 1 (~21:45) under one written mandate",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "date_text": "2026-10-01 21:06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L25"
   ]
  },
  {
   "id": "F-827",
   "kind": "event",
   "claim": "\"Off-the-shelf scan\" added to every lane (look for ready-made GitHub/HF artifacts first)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "date_text": "2026-10-01 22:45",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L27"
   ]
  },
  {
   "id": "F-828",
   "kind": "decision",
   "claim": "Blind / held-out sets: researchers never open, list or grep them; coordinator runs them by script, aggregates only — why: A filename-only grep had leaked set composition — outcome: Rule widened to blind/ on 10-10; pre-commit hook blocks blind plaintext",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 (incident) / 2026-10-10 (scope widened)",
   "source_public": null,
   "caveat": "Verification note: Same fix as the timeline row: incident at QUEUE:144; MANDATE scope widening dated 10-10.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-process-03"
   ]
  },
  {
   "id": "F-829",
   "kind": "decision",
   "claim": "Expensive strategy agent (the strategy agent, Fable tier) only for few big decisions, with a self-contained dossier; never loops — why: Cost: most expensive tier — outcome: Used once (S1 memo) in the record read",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-process-04"
   ]
  },
  {
   "id": "F-830",
   "kind": "event",
   "claim": "Bench v0.2 written by two blind writer agents (55 items each); plan frozen (combined sha fe100e91)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L31"
   ]
  },
  {
   "id": "F-831",
   "kind": "fact",
   "claim": "The strategy agent put the chance of the CI lower bound clearing 0.50 at about 1 in 4",
   "value": "~25 % (point estimate above 0.50: ~75 %)",
   "ci": "subjective",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "forecast, later wrong in the optimistic direction",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-03"
   ]
  },
  {
   "id": "F-832",
   "kind": "event",
   "claim": "Blind-set rule widened after a filename-only grep leaked set composition",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 04:45 (incident); rule widened to blind/ on 2026-10-10",
   "source_public": null,
   "caveat": "Verification note: MANDATE:104 dates the widening 2026-10-10 and does not describe the leak; the filename-only grep self-disclosure is QUEUE:144 (10-02 04:45). Do not repeat the topic word the grep revealed.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L28"
   ]
  },
  {
   "id": "F-833",
   "kind": "event",
   "claim": "Strategy agent (the strategy agent) memo S1: three fronts, ~1 in 4 chance the CI lower bound clears 0.50; stop list",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 06:10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L29"
   ]
  },
  {
   "id": "F-834",
   "kind": "decision",
   "claim": "No human review: an independent Opus reviewer agent decides content/wording; caveat \"AI-reviewed, not clinically reviewed\" stays — why: speed; author unavailable for 39 card decisions — outcome: reviewer changed 25/31 cards",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": "boar-research:reports/RULE-CHANGES.md:7",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-chron-early-17"
   ]
  },
  {
   "id": "F-835",
   "kind": "event",
   "claim": "Held-out exposure: a researcher's scan printed one v0.2 item → plaintext moved to a blind path",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": "Verification note: QUEUE:373 carries no timestamp; the exposure ledger dates the accidental print 10-03. \"22:05\" is the time of the #4b freeze at QUEUE:32, not of this incident.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-early-L63"
   ]
  },
  {
   "id": "F-836",
   "kind": "event",
   "claim": "\"No human review\": an independent Opus reviewer agent decides content; caveat \"AI-reviewed, not clinically reviewed\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": "boar-research:reports/RULE-CHANGES.md:7",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-early-L67"
   ]
  },
  {
   "id": "F-837",
   "kind": "event",
   "claim": "A research agent's trace scan prints one v0.2 item; plaintext moved to blind/, copies quarantined",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": "Verification note: EXPOSURES line is :31 (not :36). Date 10-03 is right; chron-early dated the same incident 10-02 22:05 (wrong).",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L34"
   ]
  },
  {
   "id": "F-838",
   "kind": "event",
   "claim": "Author: \"NO human review\"; an independent Opus reviewer agent decides content and wording; caveat \"AI-reviewed, not clinically reviewed\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L35"
   ]
  },
  {
   "id": "F-839",
   "kind": "event",
   "claim": "One v0.2 item printed into a research agent's trace scan; plaintext moved to an isolated folder",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": "do not name the printed item",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "T-method-L37"
   ]
  },
  {
   "id": "F-841",
   "kind": "event",
   "claim": "Overnight loop closed; 158 local commits, nothing pushed",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "date_text": "2026-10-07 08:02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L41"
   ]
  },
  {
   "id": "F-842",
   "kind": "fact",
   "claim": "Held-out sets status",
   "value": "9 sets; only one (Spanish, parked) untouched",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "source_public": "boar-research:reports/SET-STATUS.md:15-27",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-62"
   ]
  },
  {
   "id": "F-843",
   "kind": "fact",
   "claim": "65/65 criterion sha256 files verified, and the freeze commit precedes the result commit.",
   "value": "65/65",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": "ledger audit 10-13 later found 41 criteria without a sidecar sha (21 backing a decision) and one criterion edited after freezing",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-method-69"
   ]
  },
  {
   "id": "F-845",
   "kind": "fact",
   "claim": "Win-rate against the reference was 0 for every system measured (including a competitor's app)",
   "value": "0",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "V-para, v0.2",
   "set_status": null,
   "judge": "Claude Opus",
   "date": "2026-10-10",
   "source_public": null,
   "caveat": "competitor name must not be published",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-09"
   ]
  },
  {
   "id": "F-846",
   "kind": "fact",
   "claim": "The 10-10 process review",
   "value": "10 Opus agents, 76 findings; double-check: 32 claims → 10 verified, 20 partly, 2 false",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-54"
   ]
  },
  {
   "id": "F-847",
   "kind": "fact",
   "claim": "Size of the research log",
   "value": "QUEUE ≈ 51k words by 10-10; 1,303 lines by 10-17",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10/17",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-76"
   ]
  },
  {
   "id": "F-848",
   "kind": "fact",
   "claim": "Evidence lived in a single copy for days",
   "value": "rnd 136 commits ahead of origin, last push 10-06",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": "backup risk",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-77"
   ]
  },
  {
   "id": "F-849",
   "kind": "fact",
   "claim": "Pipeline efficiency of adopted levers",
   "value": "16.4 V-para floor-adj points per 1,000 judge calls (H vs D); 0 points per 1,804 calls in the 10-11/12 cycle",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para",
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10/12",
   "source_public": "boar-research:reports/BAR-TABLE.md:108,141",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-results-108"
   ]
  },
  {
   "id": "F-850",
   "kind": "fact",
   "claim": "Best-of-2 oracle over two seeds exists only with the judge in the loop",
   "value": "+0.038 raw oracle; self-selection agrees with oracle 49 % (chance)",
   "ci": "existing 2 seeds",
   "n": "51 (inferred from set)",
   "seeds": "2",
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "Opus rows",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/CLOSED.md:46",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-results-60"
   ]
  },
  {
   "id": "F-851",
   "kind": "event",
   "claim": "Double-check of that review (5 Opus agents): of 32 claims, 10 verified, 20 partly, 2 false; one task withdrawn",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 20:50",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L45"
   ]
  },
  {
   "id": "F-852",
   "kind": "event",
   "claim": "Judge pipeline made fail-closed (missing ids/shards abort; tests 33/33)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 21:16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L20"
   ]
  },
  {
   "id": "F-853",
   "kind": "event",
   "claim": "Judge pipeline made fail-closed (errored traces no longer silently dropped)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 21:16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L46"
   ]
  },
  {
   "id": "F-854",
   "kind": "decision",
   "claim": "Stop all app/claim-package work; research on the mini only until the bar is met — why: author wanted model quality raised toward the bar, not the app — outcome: HANDOFF-2026-10-11 rewritten for research-only scope; loop program followed",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 21:50–21:54",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-chron-late-02"
   ]
  },
  {
   "id": "F-855",
   "kind": "event",
   "claim": "STOP by the author (\"para tudo\"): all app/claim-package work halted",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 21:50",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L23"
   ]
  },
  {
   "id": "F-856",
   "kind": "event",
   "claim": "Criterion template with mandatory Power block; random-20 slices; bar table frozen",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 22:09",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L25"
   ]
  },
  {
   "id": "F-857",
   "kind": "decision",
   "claim": "Run the coordinator as an unattended self-paced loop; one tick = advance or start one arm, log, schedule wake-up; message the author only for a confirmed lever, weekly digest or blocker — why: Throughput without the author's attention — outcome: 133 ticks in 4 days; 1 dev lever",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-process-15"
   ]
  },
  {
   "id": "F-858",
   "kind": "decision",
   "claim": "Run research as an unattended self-paced loop (one experiment per tick, META every 8 ticks) — why: use idle machine time; keep method improving — outcome: 133 ticks over 10-11→10-14",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "date_text": "2026-10-11 03:4x",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-chron-late-04"
   ]
  },
  {
   "id": "F-859",
   "kind": "event",
   "claim": "LOOP-PROGRAM written: unattended self-paced loop, one experiment per tick",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "date_text": "2026-10-11 03:4x",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L30",
    "E-process-L48"
   ]
  },
  {
   "id": "F-860",
   "kind": "decision",
   "claim": "Measure the loss map before more levers (LOSSMAP first) — why: direct levers at the biggest loss bucket — outcome: re-ranked backlog; absent knowledge = 0.59 / 0.55 of loss",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "date_text": "2026-10-11 12:10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-chron-late-05"
   ]
  },
  {
   "id": "F-863",
   "kind": "event",
   "claim": "Deep research pass: 258 findings, 23 claims adversarially verified, 18 corrected",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L51"
   ]
  },
  {
   "id": "F-865",
   "kind": "event",
   "claim": "QUANTCEIL: 3-bit quant of deep model +0.004 (not detected); cycle adds 0 V-para points per 1,804 judge calls",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": "boar-research:reports/BAR-TABLE.md:139-142",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "T-results-L57"
   ]
  },
  {
   "id": "F-866",
   "kind": "event",
   "claim": "BAR-TABLE addendum: 0 floor-adjusted points adopted per 1,804 judge calls this cycle",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 11:27",
   "source_public": "boar-research:reports/BAR-TABLE.md:141",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L38"
   ]
  },
  {
   "id": "F-867",
   "kind": "fact",
   "claim": "One pipeline-metric reading of the 10-11→10-12 cycle",
   "value": "0 V-para floor-adj points adopted per 1,804 judge calls",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 11:27",
   "source_public": "boar-research:reports/BAR-TABLE.md:141",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-chron-late-18"
   ]
  },
  {
   "id": "F-868",
   "kind": "event",
   "claim": "RATIONALEPASS, NUMCHECK, EVALFIX, K2HALF, SEEDSPEC (0 Claude calls)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 14:17–14:32",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L40"
   ]
  },
  {
   "id": "F-869",
   "kind": "event",
   "claim": "Loop stopped by the author, restarted 12 min later on a 30-min cron",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 20:41 / 20:53",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L43"
   ]
  },
  {
   "id": "F-870",
   "kind": "decision",
   "claim": "Autonomy rule: never ask for approval; paid API only when no free equivalent, capped and logged — why: author said Claude usage is not the constraint — outcome: loop ran without approval asks",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 23:22",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-chron-late-13"
   ]
  },
  {
   "id": "F-871",
   "kind": "event",
   "claim": "Autonomy rule from the author: never ask for approval; paid API only when no free equivalent",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 23:22",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L46"
   ]
  },
  {
   "id": "F-872",
   "kind": "fact",
   "claim": "Ledger audit discrepancies",
   "value": "HIGH 12 / MEDIUM 97 / LOW 307; 48 verdicts without a closed row; 9-10 unlogged exposures",
   "ci": "54-58 Opus agents",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": "boar-research:reports/RULE-CHANGES.md:35",
   "caveat": "led to rule 28",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-11"
   ]
  },
  {
   "id": "F-873",
   "kind": "fact",
   "claim": "Closed arms by the stage at which they died",
   "value": "0-call 54 · Haiku L1 only 10 · L1 → Opus 12 · Opus direct 25 (of 101)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": null,
   "caveat": "Opus-direct mostly before Haiku L1 existed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-12"
   ]
  },
  {
   "id": "F-875",
   "kind": "event",
   "claim": "Author challenges 9+ idle ticks; loop un-idled; rules 24 (never idle while 0-call work exists) and 25 (no Haiku cap)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 08:45",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L52"
   ]
  },
  {
   "id": "F-876",
   "kind": "decision",
   "claim": "No cap on Haiku judge calls (Rule 25) — why: Haiku is cheap triage; idle caps cost machine time — outcome: 720-call scoring passes became routine",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 10:20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-chron-late-15"
   ]
  },
  {
   "id": "F-877",
   "kind": "event",
   "claim": "Rule 25: no Haiku cap",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 10:20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L51"
   ]
  },
  {
   "id": "F-878",
   "kind": "event",
   "claim": "Two session restarts; nohup jobs survived, in-process agents died",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 10:58 / 12:47",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L53"
   ]
  },
  {
   "id": "F-879",
   "kind": "event",
   "claim": "Two session restarts in one day; nohup jobs survive, in-process agents die",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 10:58, 12:47",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L53"
   ]
  },
  {
   "id": "F-880",
   "kind": "fact",
   "claim": "Dense chunk extra slots raise reach much more than quality (DENSECHUNK-ST2)",
   "value": "fact chunk in raw top-8 67 → 108; Opus +0.017",
   "ci": "(−0.000…+0.035); n 180",
   "n": "180",
   "seeds": null,
   "set": "the 180 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-13",
   "date_text": "2026-10-13 13:03",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:157",
   "caveat": "format v3 not shipped",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-chron-late-32"
   ]
  },
  {
   "id": "F-881",
   "kind": "fact",
   "claim": "MacBook 4B generation is byte-deterministic (SEEDGEN)",
   "value": "20/20 identical answers on every pairing",
   "ci": "4 regenerations",
   "n": null,
   "seeds": null,
   "set": "the 180 subset",
   "set_status": null,
   "judge": "none",
   "date": "2026-10-13",
   "date_text": "2026-10-13 13:03",
   "source_public": null,
   "caveat": "ST2's \"noise floor\" was a keying artefact; one 1/60 exception under heavy load (Rule 27a)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-chron-late-34"
   ]
  },
  {
   "id": "F-883",
   "kind": "event",
   "claim": "Ledger audit workflow (54-58 Opus agents): 159 verified arm records; 48 verdicts with no closed-registry row, 10 exposures with no row",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 21:43-22:51",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L55"
   ]
  },
  {
   "id": "F-884",
   "kind": "event",
   "claim": "Judge resume provenance fix: resuming over another trace's rows now exits 7",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 23:14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L56"
   ]
  },
  {
   "id": "F-885",
   "kind": "decision",
   "claim": "Rule 29 experiment economics: Ceiling block (launch only if ≥ 2×MDE) and Budget block (smoke, wall caps, kill rule) — why: ≈ 1,440 avoidable Opus calls and long bad tests — outcome: reviewed, edits applied",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 16:xx",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-chron-late-23"
   ]
  },
  {
   "id": "F-886",
   "kind": "decision",
   "claim": "RULE 29: Ceiling (0-call upper bound) and Budget blocks before launch; no Opus L2 on dev sets (why: ~3,400 of ~4,400 Opus calls in 3 days went to dev confirmations that matched Haiku's sign)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "outcome: in force for criteria after 10-14",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-method-25"
   ]
  },
  {
   "id": "F-887",
   "kind": "decision",
   "claim": "Verdict, registry row (CLOSED.md) and exposure row in the same commit (why: ledger audit found 48 verdicts without registry rows)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:reports/RULE-CHANGES.md:35",
   "caveat": "outcome: in force",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-method-26"
   ]
  },
  {
   "id": "F-888",
   "kind": "decision",
   "claim": "Rule 28: verdict, closed-registry row and exposure row in the same commit — why: Audit found 48 verdicts and 10 exposures unregistered — outcome: In force",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-process-20"
   ]
  },
  {
   "id": "F-889",
   "kind": "decision",
   "claim": "Rule 29 experiment economics: Ceiling block (0-call upper bound), Budget block with smoke + kill_if, no Opus on dev sets, Transfer test for dev levers — why: 10 h and 4.8 h runs wasted; ≈ 1,440 avoidable Opus calls — outcome: Enforced by template + driver; reviewer edits applied",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-process-21"
   ]
  },
  {
   "id": "F-890",
   "kind": "decision",
   "claim": "LOOP v2: machines never wait on the coordinator (queue runners with sentinel files), smoke before any run > 30 min, one STATE line per tick, handoff instead of compaction — why: Retro: 25.6 of 55.5 window-hours both machines idle; 3 restarts — outcome: Pilot 10-16: both-idle 0.7 h vs baseline 6.4 h/day",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-process-22"
   ]
  },
  {
   "id": "F-891",
   "kind": "decision",
   "claim": "Add Ceiling + Budget + Transfer blocks to every criterion; Opus only on V-para/v0.2 kit and near-misses (why: ≈ 1,440 avoidable Opus calls on dev confirmations; long runs without speed projections)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "outcome: RULE 29",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-results-14"
   ]
  },
  {
   "id": "F-892",
   "kind": "event",
   "claim": "Rule 28 (verdict + registry rows in the same commit) and Rule 29 (experiment economics: Ceiling, Budget, kill points, judge escalation); Rule 29 reviewed ACCEPT WITH EDITS",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L59"
   ]
  },
  {
   "id": "F-893",
   "kind": "fact",
   "claim": "Loop totals 2026-10-11 12:10 → 10-14 19:09: ≈ 12.0k Claude calls, of which ≈ 10.5k judge calls.",
   "value": "133 ticks (28 idle); ≈ 12.0k Claude calls = Opus 4,851 judge + Haiku 7,151 (5,660 judge + 1,491 writer/generation), i.e. ≈ 10.5k judge calls; mini 44.4 h + MacBook 18.1 h generation; 1 LEVER (dev)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "≈ 1,440 Opus calls avoidable (Haiku picked the same winner) | Verification note: Calling all 12.0k \"judge calls\" overstates judging by ≈ 1.5k; also fix in the §1 summary.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-chron-late-47"
   ]
  },
  {
   "id": "F-894",
   "kind": "fact",
   "claim": "Avoidable machine time in the loop",
   "value": "≈ 31 h (TEACHER-PROBE 10.1 mini-h, QUANTCEIL 7.1 mini-h, pre-generation 5.2, doc2query 4.8, restart 2.2, unnoticed completion 1.3)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-chron-late-48"
   ]
  },
  {
   "id": "F-895",
   "kind": "fact",
   "claim": "Long-session latency grows with history",
   "value": "≈ 22 s at turns 14–15 vs ≈ 8 s standalone (desktop); prompt +74 tokens/turn; 4,096-token context crossed at ≈ turn 50",
   "ci": "3 sessions",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": null,
   "date": "2026-10-14",
   "date_text": "10-14",
   "source_public": "boar-research:arms/longsession/LONGSESSION.md:55,80",
   "caveat": "adapter passes all turns; the app keeps 6 messages + summary from turn 8 (turns 5–7 gap)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-chron-late-57"
   ]
  },
  {
   "id": "F-896",
   "kind": "fact",
   "claim": "History helps coreference",
   "value": "coreference failures 6/20 in-session vs 17/20 standalone (5-turn run)",
   "ci": "n 20",
   "n": "20",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "agent reading",
   "date": "2026-10-14",
   "date_text": "10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-chron-late-58"
   ]
  },
  {
   "id": "F-897",
   "kind": "fact",
   "claim": "Experiment-economics evidence",
   "value": "101 closed arms: 54 at 0 calls / 10 L1 / 12 L1→L2 / 25 Opus direct; Haiku vs Opus same sign 4/4",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:reports/RULE-CHANGES.md",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-chron-late-93"
   ]
  },
  {
   "id": "F-898",
   "kind": "fact",
   "claim": "Of 101 closed arms, 54 closed at 0 judge calls, 10 at Haiku L1 only, 12 at L1 → Opus L2, 25 with Opus directly.",
   "value": "54/10/12/25",
   "ci": "101 arms",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-method-72"
   ]
  },
  {
   "id": "F-899",
   "kind": "fact",
   "claim": "Rule changes logged",
   "value": "29 numbered rules + 27a in 12 days (10-03 → 10-14)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:reports/RULE-CHANGES.md:5-37",
   "caveat": "public copy omits caps and personal quotes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-14"
   ]
  },
  {
   "id": "F-900",
   "kind": "fact",
   "claim": "Unattended loop activity 10-11 → 10-14",
   "value": "133 ticks; 28 idle (21 %); 34 event-driven; 10 META; 3 restarts",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "another doc counts 37/121 idle (different window)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-15"
   ]
  },
  {
   "id": "F-901",
   "kind": "fact",
   "claim": "Coordinator active span",
   "value": "≈ 76 h over 4 days",
   "ci": "estimate",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "estimate from first/last tick",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-16"
   ]
  },
  {
   "id": "F-903",
   "kind": "fact",
   "claim": "Machine generation time and idle time",
   "value": "mini 44.4 h, MacBook 18.1 h; both idle 25.6 of 55.5 window-hours (09-24 h)",
   "ci": "estimate",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "4 MacBook intervals estimated",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-18"
   ]
  },
  {
   "id": "F-904",
   "kind": "fact",
   "claim": "Output of the loop window",
   "value": "44 arm rows; ≈ 25 closed at step 0 (< 1 h, 0 calls); 1 LEVER",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-19"
   ]
  },
  {
   "id": "F-905",
   "kind": "fact",
   "claim": "Calls by outcome in the loop window",
   "value": "step-0 closures 1,419 · arms closed after full runs 2,972 · levers 1,896 · measurements ≈ 5,600",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-21"
   ]
  },
  {
   "id": "F-906",
   "kind": "fact",
   "claim": "Avoidable machine time",
   "value": "≈ 31 machine-hours",
   "ci": "estimate",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "reviewer estimate",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-22"
   ]
  },
  {
   "id": "F-907",
   "kind": "fact",
   "claim": "An overnight query-generation run went ahead after its own speed gate failed twice",
   "value": "4.77 MacBook-h; reach 77 vs gate 90",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "none",
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "low opportunity cost (idle night)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-24"
   ]
  },
  {
   "id": "F-910",
   "kind": "fact",
   "claim": "Incidents mined from the loop",
   "value": "26 incidents; avoidable by: 0-call precheck 8, checklist 6, deterministic hook 4, nohup script 3, smoke 2, unit test 2, write lint 1",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-41"
   ]
  },
  {
   "id": "F-911",
   "kind": "fact",
   "claim": "Session restarts and survivability",
   "value": "3 restarts in 2 days; nohup jobs survived all three; in-process agents and waiters died",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-42"
   ]
  },
  {
   "id": "F-912",
   "kind": "fact",
   "claim": "Zero-call gates were the most efficient part of the loop",
   "value": "replay gates fired 13 times and saved 13 runs; jev gates made 7 closures",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "none / jev",
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-45"
   ]
  },
  {
   "id": "F-913",
   "kind": "fact",
   "claim": "Arms and adoptions by phase",
   "value": "10-04/06: 61 arms, 12 adopted · 10-07/08: 27, 3 · 10-09/10: 22, 0 · 10-11/13: 47, 0 · 10-14: 1* (dev)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-49"
   ]
  },
  {
   "id": "F-914",
   "kind": "fact",
   "claim": "Governance overhead in ~10 days",
   "value": "8 handoffs, 5 claim-review rounds, 2 pipeline audits, 1 ledger audit with 58 agents; two coordinator sessions logging the same result twice",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-53"
   ]
  },
  {
   "id": "F-915",
   "kind": "fact",
   "claim": "How arms died",
   "value": "101 closed arms: 54 at 0 judge calls, 10 at Haiku L1, 12 at L1→Opus, 25 Opus directly",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "citation fix",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-results-109"
   ]
  },
  {
   "id": "F-916",
   "kind": "fact",
   "claim": "Multi-turn sessions do not degrade answers over 5 turns",
   "value": "turns 2–5 ratio in-session/standalone 1.397 (1.187–1.673); drift +0.537 (+0.208…+1.155)",
   "ci": "10 sessions × 5 turns; 1 seed",
   "n": null,
   "seeds": "1",
   "set": "agent-written dev",
   "set_status": "dev",
   "judge": "Haiku L1",
   "date": "2026-10-14",
   "source_public": "boar-research:arms/longsession/LONGSESSION.md:5,8,21,28",
   "caveat": "citation fix (l.3 is the health caveat). Reference = BOAR's own standalone answer, not frontier; drift is a difference of ratios",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-results-74"
   ]
  },
  {
   "id": "F-917",
   "kind": "fact",
   "claim": "Multi-turn sessions do not drift on the 4B.",
   "value": "drift (turn-5 ratio minus turn-1 ratio) +0.537 (+0.208…+1.155); every turn ≥ standalone; max prompt 928 tokens",
   "ci": "50 questions, 102 calls",
   "n": null,
   "seeds": null,
   "set": "longsession-v1 (public dev)",
   "set_status": "public dev",
   "judge": "Haiku",
   "date": "2026-10-14",
   "source_public": "boar-research:reports/CLOSED.md#R76; boar-research:reports/CLOSED.md:105; boar-research:arms/longsession/LONGSESSION.md:28",
   "caveat": "the registry calls it a 'ratio'; it is a difference of ratios vs BOAR's own standalone answers",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-system-81"
   ]
  },
  {
   "id": "F-918",
   "kind": "event",
   "claim": "RULE 29 experiment economics: Ceiling + Budget + Transfer blocks; no Opus on dev sets",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:reports/RULE-CHANGES.md:36",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "T-method-L59"
   ]
  },
  {
   "id": "F-919",
   "kind": "event",
   "claim": "EXPERIMENT-ECONOMICS: of 101 closed arms 54 closed at 0 judge calls",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "citation fix",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "T-results-L67"
   ]
  },
  {
   "id": "F-920",
   "kind": "event",
   "claim": "Ledger fixes applied: closed-registry rows R50-R70",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 01:33",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L57"
   ]
  },
  {
   "id": "F-921",
   "kind": "decision",
   "claim": "Run doc2query PoC overnight on the MacBook instead of renting a GPU — why: cost ≈ US$0.3–1 vs free local hours — outcome: 4.77 h, FAIL; later cited as an avoidable loss (speed gate overridden by an addendum)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 02:53",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-chron-late-20"
   ]
  },
  {
   "id": "F-922",
   "kind": "event",
   "claim": "MICROD2Q (doc2query) fails; doc2query closed",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 08:04",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L63"
   ]
  },
  {
   "id": "F-923",
   "kind": "fact",
   "claim": "Machine-written synthetic questions (doc2query) barely move reach",
   "value": "fact chunk top-8 69 → 77 vs 90 gate",
   "ci": "5,975 chunks, 4.77 h",
   "n": null,
   "seeds": null,
   "set": "the 180 dev",
   "set_status": "dev",
   "judge": "none",
   "date": "2026-10-14",
   "date_text": "2026-10-14 08:04",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-chron-late-46"
   ]
  },
  {
   "id": "F-924",
   "kind": "event",
   "claim": "Coordinator session dies on an expired login token; mini idle 2.2 h",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 10:4x",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L58"
   ]
  },
  {
   "id": "F-925",
   "kind": "event",
   "claim": "Author: wind-down to a decision point; no new arms",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 16:35",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L66"
   ]
  },
  {
   "id": "F-926",
   "kind": "fact",
   "claim": "Reranker on the append path (X3)",
   "value": "fact present 111 / 109 vs bar 117; 3.9 s / item",
   "ci": "n 180",
   "n": "180",
   "seeds": null,
   "set": "the 180 dev",
   "set_status": "dev",
   "judge": "none",
   "date": "2026-10-14",
   "date_text": "2026-10-14 19:09",
   "source_public": "boar-research:arms/headpass-st/X3-GATE.md",
   "caveat": "closed at step 0",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-chron-late-42"
   ]
  },
  {
   "id": "F-927",
   "kind": "decision",
   "claim": "LOOP v2 decisions D1–D12 (machine-side queue runners, smoke-first, STATE line, lanes, guards as scripts, handoff instead of compaction, daily brief, watchdog) — why: 26 incidents, ≈ 31 h avoidable machine time — outcome: T1–T7 built 10-15; T8 10-16",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 19:56",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-chron-late-24"
   ]
  },
  {
   "id": "F-928",
   "kind": "event",
   "claim": "LOOP v2 retrospective: 133 ticks, ≈ 12.0k Claude calls (≈ 10.5k judge calls), 1 lever, ≈31 avoidable machine-hours",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 19:56",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L69"
   ]
  },
  {
   "id": "F-929",
   "kind": "event",
   "claim": "LOOP v2 retrospective (10-agent workflow: 5 miners, 3 designs, 2 judges)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 19:56",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L60"
   ]
  },
  {
   "id": "F-930",
   "kind": "decision",
   "claim": "Drop the loop-operations memo and internal decision memo from the public repo; replace with a 9-line method summary — why: Reviewer: internal agent prompts, caps, people and channels — outcome: `method/METHOD.md`",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-process-25"
   ]
  },
  {
   "id": "F-931",
   "kind": "event",
   "claim": "Research page + post reviewed by an independent agent: PUBLISHABLE AFTER FIXES; every number matched the release",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L65"
   ]
  },
  {
   "id": "F-932",
   "kind": "fact",
   "claim": "Closed-family registry size",
   "value": "public 76 rows (61 closed, 13 not detected, 2 measurement); private 98 rows by 10-16",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15/16",
   "source_public": "boar-research:reports/CLOSED.md:6-9",
   "caveat": "public snapshot frozen 10-15 | Verification note: 76 = 61 + 13 + 2 is at REVIEW-RESEARCH-PAGE:5, not :33.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-13"
   ]
  },
  {
   "id": "F-933",
   "kind": "fact",
   "claim": "Research page check",
   "value": "every number matched the public release; 2 must-fix (wording)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-74"
   ]
  },
  {
   "id": "F-934",
   "kind": "fact",
   "claim": "Agent-built iOS app as a side workflow",
   "value": "10 agents, ~98 min, 2.85M subagent tokens",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": "product, not research; shows scale of agent workflows",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-75"
   ]
  },
  {
   "id": "F-935",
   "kind": "event",
   "claim": "Closing-sprint loop on in the live session",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 00:45",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L74"
   ]
  },
  {
   "id": "F-936",
   "kind": "decision",
   "claim": "First public push by the coordinator, only for the two new repos, on the author's explicit instruction — why: release — outcome: repos live; no-push rule stays for everything else",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 10:07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "D-chron-late-29"
   ]
  },
  {
   "id": "F-937",
   "kind": "event",
   "claim": "Public repos live (first push by the coordinator, explicitly authorised for those two repos only)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 10:07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L63"
   ]
  },
  {
   "id": "F-938",
   "kind": "event",
   "claim": "Engine PR opened to the upstream app; CI green 10:49",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 10:10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L64"
   ]
  },
  {
   "id": "F-939",
   "kind": "event",
   "claim": "Illustrated research page + short group post ready",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 10:48",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L78"
   ]
  },
  {
   "id": "F-940",
   "kind": "event",
   "claim": "Maintainer package sent (140-question eval request, health-path port spec)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 18:58",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L81"
   ]
  },
  {
   "id": "F-941",
   "kind": "event",
   "claim": "Session split: PR-watch session vs research session",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 19:28",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L82"
   ]
  },
  {
   "id": "F-942",
   "kind": "event",
   "claim": "LOOP v2 pilot: queue runners, one-line STATE, smoke-first gate, linted logging shipped by tooling agents",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 20:08-20:24",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L66"
   ]
  },
  {
   "id": "F-943",
   "kind": "event",
   "claim": "LOOP v2 tooling T1–T7 built (queue runner, STATE line, smoke gate, log writers, judge preflight, lane locks, watchdog)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 20:11–22:00",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L83"
   ]
  },
  {
   "id": "F-944",
   "kind": "event",
   "claim": "Coordinator's own hashing error: a valid freeze wrongly \"corrected\", then the correction corrected",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 23:38 → 10-16 00:02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L68"
   ]
  },
  {
   "id": "F-946",
   "kind": "fact",
   "claim": "LOOP v2 pilot metrics vs baseline",
   "value": "both-idle 0.7 h (of 4.1 h, 09-24) vs 6.4 h/day; useful machine-hour fraction 0.72 vs 0.63; 8.0 coordinator tool calls/tick",
   "ci": "one pilot day",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "partial day; night had 5 h idle (empty queue + login outage)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-39"
   ]
  },
  {
   "id": "F-947",
   "kind": "event",
   "claim": "Lanes idle by coordinator choice, then judge login dead (06:50); logged as a coordinator fault; \"idle lane is a failure\" restated",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 04:00-09:38",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-process-L69"
   ]
  },
  {
   "id": "F-948",
   "kind": "event",
   "claim": "Judge authentication expired; no judge passes for ≈ 2.5 h; lanes idle ≈ 5 h (coordinator fault acknowledged)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 06:50–09:38",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L89"
   ]
  },
  {
   "id": "F-950",
   "kind": "fact",
   "claim": "LOOP v2 pilot day metric (10-16)",
   "value": "both machines idle 0.7 h in 09–13 (baseline 6.4 h/day); useful machine-hour fraction 0.72",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 13:10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-chron-late-94"
   ]
  },
  {
   "id": "F-951",
   "kind": "event",
   "claim": "SMALLROUTE (small model as router) opened and closed at S0",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 18:04–18:07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L92"
   ]
  },
  {
   "id": "F-952",
   "kind": "event",
   "claim": "Independent adversarial CRITIQUE: 7 re-wordings applied",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 18:28",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "E-chron-late-L94"
   ]
  },
  {
   "id": "F-953",
   "kind": "fact",
   "claim": "Lane agents wrote technique cards",
   "value": "the retrieval research agent 19, the models/latency research agent 22, the evaluation/safety research agent 23 cards",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "date_text": "by 2026-10-17",
   "source_public": null,
   "caveat": "count of files, not of tested ideas",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-process-02"
   ]
  },
  {
   "id": "F-959",
   "kind": "fact",
   "claim": "One full v0.1 judge pass is 114 judge calls (57 items × 2 orders).",
   "value": "114 calls",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.1",
   "set_status": null,
   "judge": "Opus judge.v1",
   "date": null,
   "source_public": "boar-bench:TRACE-FORMAT.md:17",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "F-method-02"
   ]
  },
  {
   "id": "F-960",
   "kind": "null-result",
   "claim": "Learned card picker / health11 cards: detector + condition cards → wrong condition cards (altitude card on 6/8) (coordinator hand read FAIL)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-chron-early-L296"
   ]
  },
  {
   "id": "F-961",
   "kind": "null-result",
   "claim": "SYNTHDEEP: pin one source per side for comparisons → +1 item with both sides kept (gate +5) (kit lacks one side's article)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-chron-late-L266"
   ]
  },
  {
   "id": "F-962",
   "kind": "null-result",
   "claim": "NUMCHECK: answer-side check of numbers against the facts table → covers 0/16 and 0/17 wrong items (wrong numbers are of ≈ 5 kinds no single table covers)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-chron-late-L269"
   ]
  },
  {
   "id": "F-963",
   "kind": "null-result",
   "claim": "K2HALF: use 4 of 8 experts per token (llama.cpp patch) → token agreement 0.18 (bar 0.80); decode ×1.83 (quality not stable at 2-bit)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-chron-late-L270"
   ]
  },
  {
   "id": "F-964",
   "kind": "null-result",
   "claim": "Integrative decoding / multi-sample: combine several seeds → wrong facts seed-specific only 4 % (dominant pattern is one stable wrong claim)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-chron-late-L271"
   ]
  },
  {
   "id": "F-965",
   "kind": "null-result",
   "claim": "HEALTHQ: health query rewrite → 0/18 dropped destinations (not a query bug)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-chron-late-L275"
   ]
  },
  {
   "id": "F-966",
   "kind": "null-result",
   "claim": "SECTIONHIT: choose the fact's section inside kept articles → ≤ 2/12 recoverable (loss is at article level)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-chron-late-L276"
   ]
  },
  {
   "id": "F-967",
   "kind": "null-result",
   "claim": "H8-ST (contextual headers): add article first sentence as an index column → top-8 69 → 75 (bar 90) (title + section already indexed)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-chron-late-L283"
   ]
  },
  {
   "id": "F-968",
   "kind": "null-result",
   "claim": "MICROD2Q (doc2query): synthetic questions per chunk → top-8 69 → 77 (bar 90) (too small; index +18–33 %)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-chron-late-L284"
   ]
  },
  {
   "id": "F-969",
   "kind": "null-result",
   "claim": "LITE7 style swap: \"3–4 sentences + each figure\" → +0.016 (−0.035…+0.066) (flat)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-chron-late-L292"
   ]
  },
  {
   "id": "F-970",
   "kind": "null-result",
   "claim": "TK2-FIX: sources patch fix → +0.010 (−0.003…+0.024) (correctness fix, not a lever)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-chron-late-L294"
   ]
  },
  {
   "id": "F-971",
   "kind": "null-result",
   "claim": "FOLLOWUP r1 / r2: expand follow-up queries; article-anchored neighbour fetch → gold chunk 3/20 and 6/20 (bar 7) (gate miss (r2 had 0 harm))",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-chron-late-L298"
   ]
  },
  {
   "id": "F-972",
   "kind": "null-result",
   "claim": "WARN default-on: unsafe-answer warning → OOS recall 3/8; 12/18 with FP 8/92 (not reliable enough)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-chron-late-L303"
   ]
  },
  {
   "id": "F-973",
   "kind": "null-result",
   "claim": "SMALLROUTE: small model as per-item router → no signal beats chance at matched share (closed at S0 (R98))",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-chron-late-L304"
   ]
  },
  {
   "id": "F-974",
   "kind": "null-result",
   "claim": "Opus L2 on dev sets: confirm dev wins with Opus",
   "value": "≈ 3,400 Opus calls that matched Haiku's sign",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: RULE 29: no Opus L2 on dev",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-method-L221"
   ]
  },
  {
   "id": "F-975",
   "kind": "null-result",
   "claim": "\"Hardest 20\" slices: Triage on the 20 hardest items → inflated Δ (+0.091 → +0.026 full; +0.089 → +0.009) (Heterogeneity, not effect; replaced by random20)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-process-L199"
   ]
  },
  {
   "id": "F-976",
   "kind": "null-result",
   "claim": "Any Haiku Δ > 0 → Opus escalation: Rule 15 escalation → ≈ 1,440 avoidable Opus on dev; only 1 sign flip justified it (Superseded by rule 29)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-process-L200"
   ]
  },
  {
   "id": "F-978",
   "kind": "null-result",
   "claim": "Long runs without smoke: Teacher probe (10.1 h), overnight query generation (4.77 h) → NO-GO / FAIL predictable from 10-20 items (Rule 29 Budget block + LOOP v2 smoke gate)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-process-L203"
   ]
  },
  {
   "id": "F-979",
   "kind": "null-result",
   "claim": "Fresh-context `claude -p` loop ticks: Proposed replacement for the long-lived coordinator → not adopted; kept behind a shadow run (Judges ranked the warm interactive session first; T0 cost unmeasured at the time)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-process-L205"
   ]
  },
  {
   "id": "F-980",
   "kind": "null-result",
   "claim": "jev completion/dedupe checks in the loop: Calibrated classifier on deliveries and hypotheses → proposed, shadow-only (No labelled set; exact checks preferred for hard boundaries)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-process-L206"
   ]
  },
  {
   "id": "F-981",
   "kind": "null-result",
   "claim": "Cloud routines for the loop: Scheduled cloud agents → does not fit (No local files, no access to the mini or held-out sets)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-process-L207"
   ]
  },
  {
   "id": "F-982",
   "kind": "null-result",
   "claim": "Warning lexicon out of sample: WARN rule: in-sample 15/15 → out of sample health recall 3/8 (In-sample probes written next to the rule)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-process-L210"
   ]
  },
  {
   "id": "F-983",
   "kind": "null-result",
   "claim": "Win-rate as a bar conjunct: \"CI low > 0.50 AND win-rate > 0\" → 0 for every system incl. a competitor (Saturated; demoted to secondary but still reported first for honesty)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-process-L211"
   ]
  },
  {
   "id": "F-984",
   "kind": "null-result",
   "claim": "R27 Best-of-2 / self-consistency / verifier: 0 new calls",
   "value": "oracle +0.038; selectors at chance (49 %)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:46",
   "caveat": "why closed: gain needs the judge in the loop",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "agent-lab",
   "aliases": [
    "N-results-L254"
   ]
  },
  {
   "id": "F-985",
   "kind": "fact",
   "claim": "First-aid pack + A1 won all four pre-registered criteria on the iPhone profile",
   "value": "danger memory answers 6 → 1/run; Δ −0.006 (−0.029…+0.016)",
   "ci": "3 seeds, MDE 0.037",
   "n": null,
   "seeds": "3",
   "set": "v0.1 + safety track",
   "set_status": null,
   "judge": "Opus + rules",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "safety result, not a bar mover",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-early-17"
   ]
  },
  {
   "id": "F-986",
   "kind": "fact",
   "claim": "Adding Wikipedia lead sections (vital5) did nothing for the 4B",
   "value": "Δ 0.000 raw",
   "ci": "MDE 0.026, 2×2 seeds",
   "n": null,
   "seeds": "2",
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "±0.57 per-item swings hidden in a flat total (QUEUE:218)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-early-23"
   ]
  },
  {
   "id": "F-987",
   "kind": "fact",
   "claim": "The top-100k Wikipedia pack's lead vectors were mislabelled by a builder bug; fixed pack median same-article cosine 1.000 (was 0.46)",
   "value": "0.46 → 1.000",
   "ci": "40 titles",
   "n": null,
   "seeds": null,
   "set": "n/a",
   "set_status": null,
   "judge": "deterministic",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "invalidated the first pack verdicts",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-early-24"
   ]
  },
  {
   "id": "F-988",
   "kind": "fact",
   "claim": "Big-pack relevance gate failed replay: key-in-context net +3 vs bar +5, 2 items emptied, retrieval 2.17×",
   "value": "+3, 2 emptied",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.1 replay",
   "set_status": null,
   "judge": "deterministic",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "closed the knowledge front for the 4B | Verification note: Public CLOSED.md:20 (R5) is the earlier 10-01 #6 v1 gate (net +3 vs bar +6), not this 10-02 G1+ replay (net +3 vs bar +5). No public path for G1+.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-early-25"
   ]
  },
  {
   "id": "F-989",
   "kind": "fact",
   "claim": "The deep tier's gain is knowledge beyond the shipped packs (out-of-kit +0.069)",
   "value": "+0.069",
   "ci": "(+0.023…+0.114), n=41",
   "n": "41",
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "in-kit +0.036 (−0.038…+0.106), n=11",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-early-29"
   ]
  },
  {
   "id": "F-990",
   "kind": "fact",
   "claim": "v0.2-new items sit mostly outside the claim kit's knowledge packs: 72 % not-in-kit (kit A).",
   "value": "79/110",
   "ci": null,
   "n": "110 (inferred from set)",
   "seeds": null,
   "set": "v0.2",
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": "boar-bench:bench/v0.3/plan/addendum-01-coverage-split.md:19-21",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-method-64"
   ]
  },
  {
   "id": "F-991",
   "kind": "event",
   "claim": "Big-pack verdicts invalidated: pack lead vectors were mislabelled by a builder bug",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "date_text": "2026-10-02 05:25",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-early-L47"
   ]
  },
  {
   "id": "F-992",
   "kind": "fact",
   "claim": "Perfect Wikipedia retrieval buys at most ~+0.005–0.017; most oracle gain came from web-only facts",
   "value": "web-only +0.027, wiki-only +0.005",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.1 EN dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-early-48"
   ]
  },
  {
   "id": "F-993",
   "kind": "fact",
   "claim": "Big pack (top-100k Wikipedia) on the 4B: no gain, retrieval 3.3× slower",
   "value": "Δ −0.005 / −0.016",
   "ci": "1 seed",
   "n": null,
   "seeds": "1",
   "set": "v0.1 EN dev",
   "set_status": "dev",
   "judge": "Opus-med",
   "date": "2026-10-03",
   "source_public": null,
   "caveat": "one seed: under-powered (audit) | Verification note: Public CLOSED.md:20 (R5) is the 10-01 #6 v1 gate, not exp-33 (10-03). No dedicated public row for exp-33. One seed, under-powered.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-early-49"
   ]
  },
  {
   "id": "F-994",
   "kind": "fact",
   "claim": "Wikidata country-facts pack failed data review: 84 % field agreement vs 98 % bar",
   "value": "126/150",
   "ci": "30 countries × 5",
   "n": null,
   "seeds": null,
   "set": "n/a",
   "set_status": null,
   "judge": "reviewer",
   "date": "2026-10-04",
   "source_public": null,
   "caveat": "lesson: lookup packs need curated fields",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-early-58"
   ]
  },
  {
   "id": "F-995",
   "kind": "fact",
   "claim": "The original kit size.",
   "value": "3.03 GB including the 2.5 GB model; packs ≈ 0.39 GB",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-system-61"
   ]
  },
  {
   "id": "F-996",
   "kind": "fact",
   "claim": "Haiku 5.5-low vs Opus on 57 v0.1 items: same winner 54/57, r 0.922, κ 0.651, ratio −0.069 lower, 3.52 s/call, 0 failures.",
   "value": "54/57",
   "ci": "n 57, 228 calls",
   "n": "57,",
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Haiku vs Opus",
   "date": "2026-10-07",
   "source_public": "boar-research:method/L1-PROTOCOL.md:11",
   "caveat": "'same winner 54/57' carries almost no information because the reference wins ≈ 100 % of items (CRITIQUE-2026-10-16.md:32); quote the informative statistics instead: r 0.922, κ 0.651, level −0.069, Haiku Δ ≈ 0.70-0.85× the Opus Δ on pack arms",
   "verification": "OVERCLAIM",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-method-39"
   ]
  },
  {
   "id": "F-997",
   "kind": "decision",
   "claim": "Keep retrieval on the deep path (why: no-pack −0.045 (−0.083…−0.011))",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": "outcome: kept",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "D-system-12"
   ]
  },
  {
   "id": "F-998",
   "kind": "fact",
   "claim": "Without knowledge packs the deep thinking tier lost ≈ 0.045: retrieval still matters",
   "value": "Δ −0.045 (−0.083…−0.011)",
   "ci": "slice 20, 1 seed",
   "n": null,
   "seeds": "1",
   "set": "V-para dev-style",
   "set_status": "dev-style",
   "judge": "Haiku L1",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/CLOSED.md:52",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-early-100"
   ]
  },
  {
   "id": "F-999",
   "kind": "fact",
   "claim": "Adding the top-100k Wikipedia pack to the deep tier crowded out curated chunks",
   "value": "Δ −0.034 (−0.082…+0.003); 26/28 slots taken",
   "ci": "slice 20",
   "n": null,
   "seeds": null,
   "set": "V-para dev-style",
   "set_status": "dev-style",
   "judge": "Haiku L1",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/CLOSED.md:48",
   "caveat": "corpus size closed for deep tier too",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-early-105"
   ]
  },
  {
   "id": "F-1000",
   "kind": "fact",
   "claim": "Retrieval still matters for the deep tier (removing packs hurts)",
   "value": "−0.045 (−0.083…−0.011)",
   "ci": "slice 20; 1 seed",
   "n": "20 (slice)",
   "seeds": "1",
   "set": "V-para slice dev",
   "set_status": "dev",
   "judge": "Haiku N=4",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/CLOSED.md:52; boar-research:reports/CLOSED.md#R33",
   "caveat": "slice power low",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-results-63",
    "F-system-45"
   ]
  },
  {
   "id": "F-1001",
   "kind": "event",
   "claim": "NOTES35 (35B-written notes pack) fails its gate; all handoff research steps exhausted",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "date_text": "2026-10-11 02:55",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L28"
   ]
  },
  {
   "id": "F-1003",
   "kind": "decision",
   "claim": "Withdraw \"the facts are not in Wikipedia\"; no non-Wikipedia corpus build until a gate over all 34 absent V-para items runs (why: WIKISECTION scope covered only top-100k pack and 14/34 items)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": "boar-research:reports/BAR-TABLE.md:144-145; boar-research:reports/CLOSED.md:69",
   "caveat": "outcome: WIKIFULL gate → MARGINAL (9/34 verified); no WIKI-ALL arm",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "D-results-11"
   ]
  },
  {
   "id": "F-1004",
   "kind": "decision",
   "claim": "Pivot research to Knowledge Packs (product line) (why: bar levers exhausted on ≤ 12 GB; packs help users even if not the bar)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": null,
   "caveat": "outcome: 11 packs published; X2b found",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "D-results-12"
   ]
  },
  {
   "id": "F-1005",
   "kind": "decision",
   "claim": "Knowledge packs = format 2 SQLite, per-row attribution, four artefacts per pack (card, self-test, coverage probe, crowding probe) (why: zero format risk; licence compliance; measured crowding failure)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": null,
   "caveat": "outcome: 11 packs built and published 10-13",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "D-system-13"
   ]
  },
  {
   "id": "F-1006",
   "kind": "decision",
   "claim": "\"Default-eligible\" only if a pack holds ≤ 0.40 of kept slots and does not lower kit gold; otherwise user-selected only (why: crowding measured)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:33-45,50",
   "caveat": "citation fix: the rule is l.50, statuses in the table l.35-45 (5 default-eligible, 6 user-selected)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "D-system-15"
   ]
  },
  {
   "id": "F-1007",
   "kind": "fact",
   "claim": "Pipeline metric for the first loop cycle",
   "value": "0 V-para floor-adj points adopted per 1,804 judge calls (final tally of the 10-11 → 10-12 cycle); 1,180 was an interim count in BACKLOG",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "V-para",
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": "boar-research:reports/BAR-TABLE.md:141",
   "caveat": "Verification note: Cross-file: chron-late F-18 uses 1,804.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-process-50"
   ]
  },
  {
   "id": "F-1008",
   "kind": "fact",
   "claim": "Wikipedia-anywhere gate for V-para absent-knowledge items",
   "value": "9/34 verified (MARGINAL band 7–13)",
   "ci": "34 items",
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "jev + quote check",
   "date": "2026-10-12",
   "date_text": "2026-10-12/13",
   "source_public": "boar-research:reports/CLOSED.md:69",
   "caveat": "25/34 unverified, not shown absent",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-results-114"
   ]
  },
  {
   "id": "F-1009",
   "kind": "event",
   "claim": "Correction: \"facts are not in Wikipedia\" withdrawn (scope was the top-100k pack only)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": "boar-research:reports/BAR-TABLE.md:144-145",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "T-results-L58"
   ]
  },
  {
   "id": "F-1010",
   "kind": "event",
   "claim": "Peer deep-research report (258 findings, 23 claims verified, 18 corrected); \"facts are not in Wikipedia\" withdrawn",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 13:43–14:14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L39"
   ]
  },
  {
   "id": "F-1011",
   "kind": "decision",
   "claim": "Withdraw \"facts are not in Wikipedia\"; adopt the peer research's task chain — why: peer review showed WIKISECTION covered only one pack and part of the items — outcome: WIKIFULL stages run; result 9/34 (a lower bound)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 14:14",
   "source_public": null,
   "caveat": "Verification note: Outcome should read \"9/34 verified (lower bound, MARGINAL)\", not \"9/34 (a lower bound)\".",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "D-chron-late-08"
   ]
  },
  {
   "id": "F-1012",
   "kind": "decision",
   "claim": "Pivot to Knowledge Packs; primary metric = in-pack conditional ratio — why: the product is a small model + user-selected packs; the unconditional bar was not moving — outcome: P1–P4, pack workflow, pack stack",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 20:06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "D-chron-late-10"
   ]
  },
  {
   "id": "F-1013",
   "kind": "event",
   "claim": "Author's direction: Knowledge Packs; research question becomes conditional quality with the matching pack installed",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 20:06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L41"
   ]
  },
  {
   "id": "F-1014",
   "kind": "event",
   "claim": "PACKCOND P1–P4, PACKDENSE, HEALTHQ, SECTIONHIT: in-pack loss is retrieval, not grounding",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 20:14–21:15",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L42"
   ]
  },
  {
   "id": "F-1015",
   "kind": "decision",
   "claim": "Build a pack-authoring workflow; agents may write the 20 public self-tests per pack; held-out claim items stay non-Claude — why: \"we will be the first author with publications\" — outcome: 11 packs published 10-13",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 22:32",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "D-chron-late-11"
   ]
  },
  {
   "id": "F-1016",
   "kind": "event",
   "claim": "Author: build a pack-authoring workflow (\"we will be the first author with publications\")",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 22:32",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L44"
   ]
  },
  {
   "id": "F-1017",
   "kind": "decision",
   "claim": "Pack format unchanged (format 2); licence policy PD/CC0/CC BY/CC BY-SA only; never NC; per-row attribution for Stack Exchange; archive.org pre-2024 SE dumps, not the login-gated official dumps — why: zero format risk; licence compliance — outcome: applied to all published packs",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 22:46",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "D-chron-late-12"
   ]
  },
  {
   "id": "F-1018",
   "kind": "event",
   "claim": "7 wave-1 packs built",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 23:24",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L47"
   ]
  },
  {
   "id": "F-1019",
   "kind": "fact",
   "claim": "Sizes of wave-1 Q&A packs",
   "value": "travel 42,307 Q / 148.3 MB; ethereum 34,588 / 102.5 MB; bitcoin 26,042 / 77.9 MB; outdoors 5,872 / 31.6 MB; health 5,048 / 23.8 MB",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 23:24",
   "source_public": "boar-research:packs/*.md",
   "caveat": "archive.org 2024-04-02 Stack Exchange snapshot",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-22"
   ]
  },
  {
   "id": "F-1020",
   "kind": "fact",
   "claim": "Public self-tests written per pack",
   "value": "20 per pack; 140 for wave 1, 180 for 9 packs; quote/overlap checks 140/140",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "the 180, dev",
   "set_status": "dev",
   "judge": "none",
   "date": "2026-10-12",
   "date_text": "2026-10-12 23:41",
   "source_public": "boar-research:packs/selftest/",
   "caveat": "agent-written from the pack's own text: dev data, never claim items",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-24"
   ]
  },
  {
   "id": "F-1021",
   "kind": "fact",
   "claim": "Packs published on Hugging Face by 10-13 evening",
   "value": "11 packs (qa/ 7, topics/ 4) + a mirrored Vital Articles pack 10-15",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "10-13 / 10-15",
   "source_public": "boar-research:packs/README.md",
   "caveat": "licences CC BY-SA / CC0 / US PD per pack",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-21"
   ]
  },
  {
   "id": "F-1023",
   "kind": "fact",
   "claim": "Cost of the pack line's adopted dev gain",
   "value": "+0.032 at ≈ 1,840 Opus + 2,500 Haiku",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "180 self-tests (dev)",
   "set_status": "dev",
   "judge": "Haiku/Opus",
   "date": "2026-10-13",
   "source_public": null,
   "caveat": "dev only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-process-51"
   ]
  },
  {
   "id": "F-1024",
   "kind": "fact",
   "claim": "Wave-3 packs (cooking, money SE) self-test lift",
   "value": "+0.045 (+0.007…+0.087) Opus pooled",
   "ci": "n 40",
   "n": "40",
   "seeds": null,
   "set": "pack self-tests dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-13",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:48",
   "caveat": "citation fix (l.36 is the ethereum table row); Haiku CIs cross 0",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-results-115"
   ]
  },
  {
   "id": "F-1025",
   "kind": "fact",
   "claim": "Pack lift per pack tracks the keep step",
   "value": "simplewiki +0.102, ethereum-SE +0.074, bitcoin-SE +0.054; first aid +0.004; factbook −0.018",
   "ci": "20 per pack",
   "n": "180 (inferred from set)",
   "seeds": null,
   "set": "180 self-tests dev",
   "set_status": "dev",
   "judge": "Haiku",
   "date": "2026-10-13",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:9",
   "caveat": "citation fix; descriptive, 20 items per pack, Haiku",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-results-78",
    "F-system-68"
   ]
  },
  {
   "id": "F-1026",
   "kind": "fact",
   "claim": "Top-1 pack routing fails",
   "value": "accuracy 0.25 (vote) / 0.58–0.61 (4B zero-shot) vs 0.85 bar",
   "ci": "0-call / local",
   "n": "180 (inferred from set)",
   "seeds": null,
   "set": "180 self-tests",
   "set_status": "dev",
   "judge": "none",
   "date": "2026-10-13",
   "source_public": "boar-research:reports/CLOSED.md:72; boar-research:reports/CLOSED.md#R48",
   "caveat": "break-even ≈ 0.68 | break-even ≈ 0.68 accuracy",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-results-82",
    "F-chron-late-31"
   ]
  },
  {
   "id": "F-1027",
   "kind": "fact",
   "claim": "Fact chunk reaches the 4B context today",
   "value": "kept 43/180; raw top-6 55/180",
   "ci": null,
   "n": "180 (inferred from set)",
   "seeds": null,
   "set": "180 self-tests",
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": "boar-research:arms/oracle-st/ORACLE-ST.md:34",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-results-84"
   ]
  },
  {
   "id": "F-1028",
   "kind": "fact",
   "claim": "Published packs on the author's public Hugging Face dataset (boar-packs).",
   "value": "11 packs: qa/ 7, topics/ 4",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:3",
   "caveat": "the dataset URL carries the author's own handle; whether to print it is the author's decision",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-system-64"
   ]
  },
  {
   "id": "F-1029",
   "kind": "fact",
   "claim": "Pack sizes (published 11).",
   "value": "travel-SE 148.3 MB, ethereum-SE 102.5, bitcoin-SE 77.9, health-SE 23.8, outdoors-SE 31.6, firstaid 21.0, medlineplus 4.3, factbook 11.6, simplewiki 480.5, cooking-SE 84.9 MiB, money-SE 130.9 MiB",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": "boar-research:packs/README.md:15-25",
   "caveat": "cooking/money 84.9 / 130.9 MiB = 89.0 / 137.2 MB (REPORT-PACKS l.44-45)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-system-65"
   ]
  },
  {
   "id": "F-1030",
   "kind": "fact",
   "claim": "Pack licences.",
   "value": "Stack Exchange CC BY-SA 2.5/3.0/4.0 per post (archive.org dump 2024-04-02); simplewiki CC BY-SA 4.0 (+GFDL); factbook CC0; MedlinePlus public domain (credit line); firstaid CC BY-SA + US PD",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": "boar-research:packs/README.md:15-25; boar-research:reports/REPORT-PACKS-2026-10-13.md:35-45",
   "caveat": "SE dump is 2.5 years stale",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-system-66"
   ]
  },
  {
   "id": "F-1031",
   "kind": "fact",
   "claim": "One matching pack lifts the 4B.",
   "value": "0.460 → 0.493; Δ +0.032 (+0.016…+0.050) Haiku; +0.042 (+0.024…+0.062) Opus; win 0/180",
   "ci": "n 180, 1 seed",
   "n": "180,",
   "seeds": "1",
   "set": "180 pack self-tests (dev, agent-written)",
   "set_status": "dev",
   "judge": "Haiku L1 / Opus L2",
   "date": "2026-10-13",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:8",
   "caveat": "self-tests favour the pack",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-system-67"
   ]
  },
  {
   "id": "F-1032",
   "kind": "fact",
   "claim": "The 35B deep tier gains nothing from packs.",
   "value": "0.546 vs 0.548, Δ −0.002 (−0.020…+0.016); deep vs 4B with same pack +0.054 (+0.034…+0.073)",
   "ci": "n 180",
   "n": "180",
   "seeds": null,
   "set": "180 self-tests (dev)",
   "set_status": "dev",
   "judge": "Haiku",
   "date": "2026-10-13",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:10",
   "caveat": "contrast with F-system-45 (retrieval matters on V-para)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-system-69"
   ]
  },
  {
   "id": "F-1033",
   "kind": "fact",
   "claim": "Top-1 pack routing fails.",
   "value": "deterministic vote 0.25; 4B zero-shot 0.58/0.61 vs 0.85 bar; break-even ≈ 0.68",
   "ci": "n 180",
   "n": "180",
   "seeds": null,
   "set": "180 self-tests",
   "set_status": null,
   "judge": "none",
   "date": "2026-10-13",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:13",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-system-71"
   ]
  },
  {
   "id": "F-1034",
   "kind": "event",
   "claim": "Knowledge packs published on Hugging Face: 7 at 00:01 and 9 at 02:30 on 10-13, 11 by 19:34 the same day; single pack +0.042 Opus on its 180 self-tests; deep tier gains nothing from packs (−0.002)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:3,8,10",
   "caveat": "'9 packs' is a mid-day count; system notes and the public report say 11. REPORT-PACKS l.7 is blank: numbers are at l.8 and l.10",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "T-results-L59"
   ]
  },
  {
   "id": "F-1035",
   "kind": "event",
   "claim": "Pack-stack technical report (crowding, routing, keep, reader)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:1-23",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "T-system-L56"
   ]
  },
  {
   "id": "F-1036",
   "kind": "decision",
   "claim": "Publish packs that pass the four-artefact check (card, self-test, coverage, crowding) on HF under the author's standing OK — why: public reuse; licence clarity per pack — outcome: 7 + 2 + 2 packs published",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 00:01",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "D-chron-late-14"
   ]
  },
  {
   "id": "F-1037",
   "kind": "event",
   "claim": "Wave 1 (7 packs) published on Hugging Face (the author's boar-packs dataset)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 00:01",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L48"
   ]
  },
  {
   "id": "F-1038",
   "kind": "event",
   "claim": "Wave 1 packs published on Hugging Face (7 packs)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 00:01",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "T-system-L53"
   ]
  },
  {
   "id": "F-1039",
   "kind": "event",
   "claim": "WIKIFULL (all of Wikipedia + 2026-10-01 delta) holds the missing fact for 9/34 (a lower bound) V-para absent items: Wikipedia closed as kit content for the bar",
   "value": "WIKIFULL stage A+B: 9/34 V-para absent items verified (a lower bound, MARGINAL band 7-13; REOPEN needed ≥ 14); unfetched pages could add about 0-3; Wikipedia not reopened as kit content",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 01:50",
   "source_public": null,
   "caveat": "Verification note: \"9/34 (a lower bound)\" inverts the bound: 9/34 is a LOWER bound; the source reads \"about a quarter at most\".",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L49"
   ]
  },
  {
   "id": "F-1040",
   "kind": "fact",
   "claim": "Wave-2 pack sizes",
   "value": "Factbook 261 entries / 11.6 MB; Simple English Wikipedia 208,622 articles / 480.5 MB",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 02:16",
   "source_public": "boar-research:packs/boar-factbook.md,boar-simplewiki-en.md",
   "caveat": "CIA Factbook discontinued 2026-02-04; CC0 archive",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-23"
   ]
  },
  {
   "id": "F-1041",
   "kind": "event",
   "claim": "Wave 2 (Factbook, Simple English Wikipedia) published",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 02:30",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L50"
   ]
  },
  {
   "id": "F-1042",
   "kind": "event",
   "claim": "SELFTEST-SCORES: 9 packs scored with/without on the 180",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 10:37",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L52"
   ]
  },
  {
   "id": "F-1043",
   "kind": "fact",
   "claim": "One matching pack lifts the 4B on its own self-tests",
   "value": "raw 0.460 → 0.493, Δ +0.032",
   "ci": "(+0.016…+0.050); n 180; 1 seed; better/worse 93/50; win 0/180",
   "n": "180",
   "seeds": "1",
   "set": "the 180 dev",
   "set_status": "dev",
   "judge": "Haiku L1",
   "date": "2026-10-13",
   "date_text": "2026-10-13 10:37",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:69",
   "caveat": "per pack: simplewiki +0.102, ethereum +0.074, bitcoin +0.054",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-26"
   ]
  },
  {
   "id": "F-1044",
   "kind": "event",
   "claim": "Pack-stack measurement: DENSECHUNK-ST2, PACKROUTE, PACKROUTE-2, GROUND-ST, CROWDQ, STACK/STACK-2, SEEDGEN",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 12:33–17:08",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L54"
   ]
  },
  {
   "id": "F-1045",
   "kind": "decision",
   "claim": "Run an oracle/ceiling cell (ORACLE-ST) before more pre-processing levers — why: decide whether reader or retrieval is the wall — outcome: ceiling +0.095 found; became a standing rule (LOOP v2 D8)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 13:4x",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "D-chron-late-16"
   ]
  },
  {
   "id": "F-1046",
   "kind": "decision",
   "claim": "Routing among packs closed at the product level; keep rule + quota only an optional setting — why: crowding cost small (−0.016 Haiku / −0.019 Opus); routers 0.25–0.61 accuracy — outcome: product reading: ship small user-selected packs, no automatic routing",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 15:44",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "D-chron-late-17"
   ]
  },
  {
   "id": "F-1047",
   "kind": "fact",
   "claim": "Combined keep rule + quota under many packs (STACK-2)",
   "value": "Δ +0.005",
   "ci": "(−0.009…+0.021); n 180",
   "n": "180",
   "seeds": null,
   "set": "the 180 dev",
   "set_status": "dev",
   "judge": "Haiku L1",
   "date": "2026-10-13",
   "date_text": "2026-10-13 17:08",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:196",
   "caveat": "not decided at MDE 0.03; Opus +0.004",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-33"
   ]
  },
  {
   "id": "F-1048",
   "kind": "event",
   "claim": "REPORT-PACKS drafted and independently reviewed (39 edits)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 17:15–17:21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L55"
   ]
  },
  {
   "id": "F-1049",
   "kind": "event",
   "claim": "Wave 3 (cooking, money SE) published; hub holds 11 packs",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 19:34",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L56"
   ]
  },
  {
   "id": "F-1050",
   "kind": "event",
   "claim": "Wave 3 published (cooking, money); hub holds 11 packs",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 19:34",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "T-system-L55"
   ]
  },
  {
   "id": "F-1051",
   "kind": "event",
   "claim": "SELFTEST-DEEP 18/18: the deep tier gains nothing from any pack",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 19:36",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L57"
   ]
  },
  {
   "id": "F-1052",
   "kind": "fact",
   "claim": "The deep tier gains nothing from any pack (SELFTEST-DEEP)",
   "value": "with 0.546 vs without 0.548, Δ −0.002",
   "ci": "(−0.020…+0.016); n 180; 18 runs",
   "n": "180",
   "seeds": null,
   "set": "the 180 dev",
   "set_status": "dev",
   "judge": "Haiku L1",
   "date": "2026-10-13",
   "date_text": "2026-10-13 19:36",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:98",
   "caveat": "MedlinePlus −0.049 (−0.104…−0.002)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-28"
   ]
  },
  {
   "id": "F-1055",
   "kind": "event",
   "claim": "ORACLE-ST: injecting the known fact chunk = +0.095 → retrieval reach is the wall on pack items, not the reader",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 23:10",
   "source_public": "boar-research:arms/oracle-st/ORACLE-ST.md:3",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "T-results-L60"
   ]
  },
  {
   "id": "F-1056",
   "kind": "event",
   "claim": "ORACLE-ST: one injected fact chunk +0.095 → retrieval reach is the headroom",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "date_text": "2026-10-13 23:11",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L60"
   ]
  },
  {
   "id": "F-1057",
   "kind": "fact",
   "claim": "Diagnosis of the 137 items whose fact chunk never reached the context",
   "value": "article found but wrong chunk 56; relevance gate 21; keep drops 20; crowding ≤ 12; lexical misses 27",
   "ci": "n 137",
   "n": "137",
   "seeds": null,
   "set": "the 180 dev",
   "set_status": "dev",
   "judge": "none",
   "date": "2026-10-13",
   "date_text": "2026-10-13 23:49",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-36"
   ]
  },
  {
   "id": "F-1058",
   "kind": "decision",
   "claim": "Bounty chat gets method, packs and device proof only; numbers only after the shared rubric exists — why: numbers without a shared rubric are not comparable — outcome: post drafts carry no scores",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "D-chron-late-28"
   ]
  },
  {
   "id": "F-1059",
   "kind": "decision",
   "claim": "X2b (neighbour + budgeted dense append) recommended as pack-line default when a format-v3 pack is installed (why: LEVER by frozen rule +0.066 Opus; no harm on kit)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:arms/headpass-st/HEADPASS-ST.md:89-91",
   "caveat": "outcome: Ported to lite tier (LITE5); not a bar lever",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "D-results-13"
   ]
  },
  {
   "id": "F-1060",
   "kind": "fact",
   "claim": "X2b costs",
   "value": "prompt tokens p50 453 → 872; kit byte-identical; v0.1 + simplewiki no-harm +0.012",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "the 180 / v0.1 public",
   "set_status": null,
   "judge": "Opus",
   "date": "2026-10-14",
   "date_text": "10-14",
   "source_public": null,
   "caveat": "needs format-v3 sidecars (+≈ 60 % pack size)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-40"
   ]
  },
  {
   "id": "F-1061",
   "kind": "fact",
   "claim": "≈ 3,400 of ≈ 4,400 Opus calls on 10-12 → 10-14 went to dev-set confirmations (self-tests) that matched Haiku's sign.",
   "value": "3,400 / 4,400",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-14",
   "source_public": "boar-research:reports/RULE-CHANGES.md:36",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-method-73"
   ]
  },
  {
   "id": "F-1063",
   "kind": "fact",
   "claim": "X2b conversion where the fact chunk is newly present",
   "value": "+0.177 (+0.098…+0.263)",
   "ci": "n 30",
   "n": "30",
   "seeds": null,
   "set": "180 self-tests dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-14",
   "source_public": "boar-research:arms/headpass-st/HEADPASS-ST.md:59",
   "caveat": "appending non-fact chunks ≈ 0 cost (+0.008) | —",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-results-87",
    "F-chron-late-39"
   ]
  },
  {
   "id": "F-1064",
   "kind": "fact",
   "claim": "X2b reach and cost",
   "value": "fact present 102/180 (S 35); +≈ 420 prompt tokens p50; total p50 13.4 s vs 8.3 s; sidecars ≈ +60 % pack size",
   "ci": null,
   "n": "180 (inferred from set)",
   "seeds": null,
   "set": "180",
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:arms/headpass-st/HEADPASS-ST.md:51,89-91",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-results-88"
   ]
  },
  {
   "id": "F-1065",
   "kind": "event",
   "claim": "HEADPASS-X2b lever (+0.066 on 180 dev self-tests)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:21",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "T-system-L57"
   ]
  },
  {
   "id": "F-1066",
   "kind": "event",
   "claim": "HEADPASS X1: fact chunk in context 35 → 72/180; Opus +0.033 (label capped)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 01:47–03:00",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L62"
   ]
  },
  {
   "id": "F-1067",
   "kind": "event",
   "claim": "HEADPASS X1 (neighbour chunks): Opus +0.033 (measured beyond plan; label capped \"not decidable\")",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 02:4x",
   "source_public": "boar-research:arms/headpass-st/HEADPASS-ST.md:45",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "T-results-L62"
   ]
  },
  {
   "id": "F-1068",
   "kind": "event",
   "claim": "HEADPASS-X2b: fact present 102/180; Opus +0.066 → first LEVER of the pack line",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 14:14–15:38",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L65"
   ]
  },
  {
   "id": "F-1069",
   "kind": "fact",
   "claim": "HEADPASS-X2b (X1 + budgeted dense candidates, +600 tokens per item)",
   "value": "fact present 102/180; Opus +0.066; Haiku +0.056",
   "ci": "(+0.043…+0.091) / (+0.036…+0.076); n 180; 1 seed",
   "n": "180",
   "seeds": "1",
   "set": "the 180 dev",
   "set_status": "dev",
   "judge": "Opus + Haiku",
   "date": "2026-10-14",
   "date_text": "2026-10-14 15:30",
   "source_public": "boar-research:arms/headpass-st/HEADPASS-ST.md:57,89",
   "caveat": "LEVER by frozen rule on pack-matched dev only; designed on the same 180",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-38"
   ]
  },
  {
   "id": "F-1070",
   "kind": "decision",
   "claim": "HEADPASS-X2b = the pack line's retrieval recommendation for format-v3 packs; X1 for format-2 packs — why: Opus +0.066 (+0.043…+0.091), no harm on kit — outcome: X3 failed at step 0; X2b staged for after 10-20 (PR-2)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 15:38",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "D-chron-late-21"
   ]
  },
  {
   "id": "F-1071",
   "kind": "event",
   "claim": "HEADPASS-X2b = LEVER on 180 pack-matched dev items: +0.066 (+0.043…+0.091) Opus",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 17:1x",
   "source_public": "boar-research:arms/headpass-st/HEADPASS-ST.md:89",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "T-results-L66"
   ]
  },
  {
   "id": "F-1072",
   "kind": "fact",
   "claim": "Detector v3b dev numbers at threshold 0.2.",
   "value": "curated recall 179/180, strict false fires 15/100, ES 54/54, 0 critical misses; realistic recall 240/246, false fires 3/457",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "dev sets (threshold not tuned on them)",
   "set_status": "dev",
   "judge": "none (deterministic)",
   "date": "2026-10-15",
   "source_public": null,
   "caveat": "dev data",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-system-05"
   ]
  },
  {
   "id": "F-1073",
   "kind": "fact",
   "claim": "HEALTHGAP predicate no-harm.",
   "value": "fires 0/60 v0.1, 1/220 self-tests",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "public dev",
   "set_status": "public dev",
   "judge": "none",
   "date": "2026-10-15",
   "source_public": null,
   "caveat": "citation fix (l.163 is the 15/64 sentence)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-system-14"
   ]
  },
  {
   "id": "F-1074",
   "kind": "fact",
   "claim": "Health-path footprint for a port.",
   "value": "detector 106.2 MB + 1.7 MB + 9 KB; cards 76 KB; firstaid pack 20.0 MiB",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": "citation fix",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-system-15"
   ]
  },
  {
   "id": "F-1075",
   "kind": "fact",
   "claim": "Ternary 2B vs stock Qwen3-1.7B (LITE2)",
   "value": "kit −0.057; self-tests −0.020",
   "ci": "(−0.104…−0.014) / (−0.077…+0.038); n 20 each",
   "n": "20",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Haiku L1",
   "date": "2026-10-15",
   "date_text": "2026-10-15 23:14",
   "source_public": null,
   "caveat": "1.7B vs 4B −0.111 on the kit",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-66"
   ]
  },
  {
   "id": "F-1076",
   "kind": "fact",
   "claim": "X2b on the claim kit (bounty-shaped)",
   "value": "inert: kept/prompt identical 20/20; KIT-X2b closed (R93)",
   "ci": "n 20",
   "n": "20",
   "seeds": null,
   "set": "v0.2 random20 dev",
   "set_status": "dev",
   "judge": "none",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "not a bounty-bar lever",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-41"
   ]
  },
  {
   "id": "F-1077",
   "kind": "fact",
   "claim": "Critique: multiplicity is not the main problem; selection on the same dev sets is",
   "value": "≈ 60 judged comparisons, ≈ 5 lever-class; X2b z ≈ 5.5",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-89"
   ]
  },
  {
   "id": "F-1078",
   "kind": "fact",
   "claim": "Multiplicity check",
   "value": "≈ 60 judged comparisons → ≈ 1.5 false \"CI low > 0\" expected; X2b z ≈ 5.5 survives Bonferroni over 100",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "main risk is selection on dev sets, not multiplicity",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-process-60"
   ]
  },
  {
   "id": "F-1079",
   "kind": "fact",
   "claim": "Large levers survive multiplicity correction",
   "value": "X2b z ≈ 5.5, LITE5-180 z ≈ 6, DEEPTHINK z ≈ 4.2",
   "ci": "≈ 60 judged comparisons",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "citation fix; ≈ 60 judged comparisons, Bonferroni over 100; the bigger risk is selection on the same dev sets",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-results-111"
   ]
  },
  {
   "id": "F-1080",
   "kind": "fact",
   "claim": "X2b on the lite 1.7B",
   "value": "+0.071 (+0.048…+0.094)",
   "ci": "n 180; 1 seed",
   "n": "180",
   "seeds": "1",
   "set": "180 self-tests dev",
   "set_status": "dev",
   "judge": "Haiku",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-results-90"
   ]
  },
  {
   "id": "F-1081",
   "kind": "fact",
   "claim": "1.7B + X2b vs plain 4B on pack-matched items",
   "value": "−0.007 (−0.031…+0.018)",
   "ci": "n 180",
   "n": "180",
   "seeds": null,
   "set": "180 self-tests dev",
   "set_status": "dev",
   "judge": "Haiku",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "does NOT hold on the kit: 1.7B − 4B −0.111 (−0.191…−0.047)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-results-91"
   ]
  },
  {
   "id": "F-1082",
   "kind": "fact",
   "claim": "Lite gap persists with identical contexts",
   "value": "1.7B X2b − 4B X2b −0.070 (−0.094…−0.044)",
   "ci": "n 97",
   "n": "97",
   "seeds": null,
   "set": "180 self-tests dev",
   "set_status": "dev",
   "judge": "Haiku",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "contradicts \"gap is reach, not reading\" | contradicts the earlier \"reach, not reading\" summary",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-results-92",
    "F-chron-late-68"
   ]
  },
  {
   "id": "F-1083",
   "kind": "fact",
   "claim": "Lite tier (Qwen3-1.7B Q4_K_M) footprint and speed.",
   "value": "file 1,107,409,472 B; RSS ≈ 2.3 GB (peak 2.7) vs ≈ 4.0 GB 4B; desktop prefill 2.5×, decode 2.2× the 4B",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "≈ 1.2× faster with X2b at matched quality; phone unpriced",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-system-78"
   ]
  },
  {
   "id": "F-1084",
   "kind": "fact",
   "claim": "Lite tier quality, correctly scoped.",
   "value": "1.7B + X2b vs plain 4B −0.007 (−0.031…+0.018) on 180 self-tests; on the claim kit 1.7B − 4B = −0.111 and X2b inert; at identical contexts −0.070 (−0.094…−0.044)",
   "ci": "n 180 / n 97",
   "n": "180",
   "seeds": null,
   "set": "180 self-tests (dev) / claim kit",
   "set_status": "dev",
   "judge": "Haiku",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "\"1.7B = 4B\" holds only on pack-matched dev items",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-system-79"
   ]
  },
  {
   "id": "F-1086",
   "kind": "event",
   "claim": "LITE5-180: X2b transfers to the 1.7B (+0.071 Haiku)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 00:56",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "E-chron-late-L88"
   ]
  },
  {
   "id": "F-1087",
   "kind": "fact",
   "claim": "X2b on the 1.7B (LITE5-180)",
   "value": "+0.071; 1.7B+X2b − 4B plain −0.007",
   "ci": "(+0.048…+0.094) / (−0.031…+0.018); n 180",
   "n": "180",
   "seeds": null,
   "set": "the 180 dev",
   "set_status": "dev",
   "judge": "Haiku L1",
   "date": "2026-10-16",
   "date_text": "2026-10-16 00:56",
   "source_public": null,
   "caveat": "\"= 4B\" holds only on pack-matched dev; on the kit 1.7B −0.111 and X2b inert",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-67"
   ]
  },
  {
   "id": "F-1088",
   "kind": "event",
   "claim": "LITE5-180: X2b on 1.7B +0.071 (Haiku); 1.7B+X2b = 4B plain (−0.007) on pack-matched items",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 00:5x",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "T-results-L70"
   ]
  },
  {
   "id": "F-1089",
   "kind": "fact",
   "claim": "Duplicate sources in answers",
   "value": "repeated page in 36–50 % of sourced answers; 90 % with X2b",
   "ci": "572 traces",
   "n": null,
   "seeds": null,
   "set": "public sets",
   "set_status": null,
   "judge": "none",
   "date": "2026-10-16",
   "date_text": "2026-10-16 11:23",
   "source_public": null,
   "caveat": "cosmetic; grouping patch staged",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-95"
   ]
  },
  {
   "id": "F-1090",
   "kind": "fact",
   "claim": "Lite speed",
   "value": "1.7B/4B prefill 2.5×, decode 2.2× (mini CPU); end-to-end with X2b ≈ 1.2× desktop",
   "ci": "10 prompts",
   "n": null,
   "seeds": null,
   "set": "synthetic",
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 18:46",
   "source_public": null,
   "caveat": "phone unmeasured",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-chron-late-69"
   ]
  },
  {
   "id": "F-1091",
   "kind": "fact",
   "claim": "Retrieval and the similarity gate use bge-small-en-v1.5 q8_0 embeddings (384 dims), shared by every pack.",
   "value": "36,806,944 B, MIT",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "date_text": "2026-10",
   "source_public": "boar-research:BOM.md:10; boar-research:packs/README.md:29",
   "caveat": "private README line is 27 (29 is the public copy)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-system-02"
   ]
  },
  {
   "id": "F-1092",
   "kind": "fact",
   "claim": "Places pack example from the upstream app.",
   "value": "Berlin 3.3 MB for 15,278 places",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "date_text": "2026-10",
   "source_public": null,
   "caveat": "maintainer's figure",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "F-system-77"
   ]
  },
  {
   "id": "F-1093",
   "kind": "null-result",
   "claim": "Big-pack (top-100k) for the 4B (#6/#6b/#17): pack, named cap, relevance gate → net +3 keys, 2 emptied, 2–3× slower (gates failed; front closed (K1))",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-early-L286"
   ]
  },
  {
   "id": "F-1094",
   "kind": "null-result",
   "claim": "Big-pack title tier (exp-33): top-100k as title lookup → Δ −0.005 / −0.016 (KILL (1 seed))",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-early-L298"
   ]
  },
  {
   "id": "F-1095",
   "kind": "null-result",
   "claim": "Country-facts pack (CF): Wikidata lookup → 84 % field agreement (bar 98 %) (data quality)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-early-L305"
   ]
  },
  {
   "id": "F-1096",
   "kind": "null-result",
   "claim": "Crypto EIP selection (CRYPTOSEL): in-pack EIP selection → targets +0.47 (bar +0.5) (FAIL)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-early-L307"
   ]
  },
  {
   "id": "F-1097",
   "kind": "null-result",
   "claim": "Prompt diet (PROMPTDIET/SPEEDPACK): −126 fixed tokens + 96-token cap → Δ −0.016; CW 5 → 9 (quality fell)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:22",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-early-L308"
   ]
  },
  {
   "id": "F-1098",
   "kind": "null-result",
   "claim": "Retrieval removal on deep tier (DEEPNORAG): no packs → −0.045 (packs stay (diagnostic))",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-early-L329"
   ]
  },
  {
   "id": "F-1099",
   "kind": "null-result",
   "claim": "Corpus size on deep tier (DEEPWIKI): top-100k pack → −0.034 (crowding)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-early-L334"
   ]
  },
  {
   "id": "F-1100",
   "kind": "null-result",
   "claim": "DEEPCOV-FB: Wikipedia pack as a fallback when nothing on-topic is kept → gold enters on 1 item (gate 5) (fourth retrieval-side NO on the deep tier)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-late-L260"
   ]
  },
  {
   "id": "F-1101",
   "kind": "null-result",
   "claim": "NOTES35: 35B writes 300 topic notes as a pack → note kept on 3/110 (gate 5 %) (curated chunks win the slots)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-late-L261"
   ]
  },
  {
   "id": "F-1102",
   "kind": "null-result",
   "claim": "VITALPACK-DEEP: Vital Articles intros pack as a peer → crowding 61 % of kept slots on v0.2 (gate 40 %) (crowding)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-late-L262"
   ]
  },
  {
   "id": "F-1103",
   "kind": "null-result",
   "claim": "WIKISECTION (H5/H5b): is the missing fact in the kept Wikipedia article body? → 2/13 V-para, 4/30 v0.2 (bar 40 %) (scope later corrected; superseded by WIKIFULL)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-late-L267"
   ]
  },
  {
   "id": "F-1104",
   "kind": "null-result",
   "claim": "PACKCOND P4 / packGround: relax grounding for pack passages → 14/32 (bar 16); changes 6/20 controls (bar 4) (loss is retrieval, not grounding)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-late-L273"
   ]
  },
  {
   "id": "F-1105",
   "kind": "null-result",
   "claim": "PACKDENSE: lead-vector dense router within packs → 21/32 vs 22 (bge-small leads do not separate travel guides)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-late-L274"
   ]
  },
  {
   "id": "F-1106",
   "kind": "null-result",
   "claim": "PACKROUTE / PACKROUTE-2: deterministic vote router; 4B zero-shot router → accuracy 0.25; 0.58/0.61 (bar 0.85) (a wrong route loses everything)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-late-L280"
   ]
  },
  {
   "id": "F-1107",
   "kind": "null-result",
   "claim": "GROUND-ST: keep-step rules for pack passages → fact chunk kept 54 vs 55 bar (recovers articles, not chunks)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-late-L281"
   ]
  },
  {
   "id": "F-1108",
   "kind": "null-result",
   "claim": "STACK / STACK-2: keep rule + per-pack quota under many packs → +0.005 (−0.009…+0.021) (not decided; optional setting only)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-late-L282"
   ]
  },
  {
   "id": "F-1109",
   "kind": "null-result",
   "claim": "HEADPASS-X2 (unbudgeted): append dense candidates with no token budget → failed prompt-token gate by 20.5 tokens (replaced by budgeted X2b)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-late-L286"
   ]
  },
  {
   "id": "F-1110",
   "kind": "null-result",
   "claim": "HEADPASS-X3: cross-encoder reranker picks appended chunks → 111/109 vs 117; 3.9 s/item (slower and below the bar)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-late-L287"
   ]
  },
  {
   "id": "F-1111",
   "kind": "null-result",
   "claim": "LITEREACH2: double the X2b budget on the 1.7B → +0.005 (−0.014…+0.025) on the 180 (more context, no gain)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-late-L297"
   ]
  },
  {
   "id": "F-1112",
   "kind": "null-result",
   "claim": "HEALTHFALLBACK / HEALTHDROP: deep answer or pack-source rescue when health route has no source → rescue 1/18; 10 wrong-topic rescues (overlap with WARN/HEALTHGAP; harm)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-late-L301"
   ]
  },
  {
   "id": "F-1113",
   "kind": "null-result",
   "claim": "KIT-X2b: X2b on the claim kit → gold-in-kept 15 → 15 (inert without v3 packs)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-chron-late-L302"
   ]
  },
  {
   "id": "F-1114",
   "kind": "null-result",
   "claim": "R5 enwiki top-100k pack for 4B: key-in-context gate",
   "value": "+3 (bar +6); retrieval 3.3× slower",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:20",
   "caveat": "why closed: gate FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L232"
   ]
  },
  {
   "id": "F-1115",
   "kind": "null-result",
   "claim": "R29 enwiki top-100k as peer pack (deep): slice 20",
   "value": "−0.034 (−0.082…+0.003)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:48",
   "caveat": "why closed: crowding measured",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L256"
   ]
  },
  {
   "id": "F-1116",
   "kind": "null-result",
   "claim": "R32 Wiki pack as fallback: replay",
   "value": "gold enters on 1 item (gate 5)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:51",
   "caveat": "why closed: gate FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L259"
   ]
  },
  {
   "id": "F-1117",
   "kind": "null-result",
   "claim": "R33 Remove retrieval from deep tier: slice 20",
   "value": "−0.045 (−0.083…−0.011)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:52",
   "caveat": "why closed: diagnostic: packs stay",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L260"
   ]
  },
  {
   "id": "F-1118",
   "kind": "null-result",
   "claim": "R34 Vital Articles intros as peer pack: replay",
   "value": "pack share 0.61 (bar ≤ 0.40)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:53",
   "caveat": "why closed: gate FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L261"
   ]
  },
  {
   "id": "F-1119",
   "kind": "null-result",
   "claim": "R39 Wikipedia sections for absent items: jev gate",
   "value": "fact in body 2/13, 4/30 (bar 40 %)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:63",
   "caveat": "why closed: step-0 FAIL (scope limited; superseded by R45)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L266"
   ]
  },
  {
   "id": "F-1120",
   "kind": "null-result",
   "claim": "R44 0-LLM recall levers on installed packs: simulation",
   "value": "21/32 vs gate 22",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:68",
   "caveat": "why closed: step-0 FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L271"
   ]
  },
  {
   "id": "F-1121",
   "kind": "null-result",
   "claim": "R45 All of Wikipedia as kit content: fact-anywhere gate",
   "value": "9/34 verified",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:69",
   "caveat": "why closed: MARGINAL (no arm)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L272"
   ]
  },
  {
   "id": "F-1122",
   "kind": "null-result",
   "claim": "R46 Dense chunk extra slots: Opus on 180",
   "value": "+0.017 (−0.000…+0.035)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:70",
   "caveat": "why closed: not decided at MDE 0.03; parked",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L273"
   ]
  },
  {
   "id": "F-1123",
   "kind": "null-result",
   "claim": "R48 Top-1 pack routing: vote + 4B classifier",
   "value": "0.25 / 0.58 / 0.61 vs 0.85",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:72",
   "caveat": "why closed: step-0 FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L275"
   ]
  },
  {
   "id": "F-1124",
   "kind": "null-result",
   "claim": "R49 Combined pack stack: Haiku on 180",
   "value": "+0.005 (−0.009…+0.021)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:73",
   "caveat": "why closed: not decided; optional setting",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L276"
   ]
  },
  {
   "id": "F-1125",
   "kind": "null-result",
   "claim": "R64 Travel-pack selection: L0 gold gate",
   "value": "1/3 (bar 2)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:93",
   "caveat": "why closed: FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L285"
   ]
  },
  {
   "id": "F-1126",
   "kind": "null-result",
   "claim": "R73 Unbudgeted dense append (X2): replay",
   "value": "reach PASS but +20.5 tokens over cost gate",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md:102",
   "caveat": "why closed: FAIL → fixed by X2b (LEVER)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L294"
   ]
  },
  {
   "id": "F-1127",
   "kind": "null-result",
   "claim": "R79 X2b on lite 1.7B: 180 Haiku",
   "value": "+0.071 (+0.048…+0.094)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: LEVER (registered, not a null)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L300"
   ]
  },
  {
   "id": "F-1128",
   "kind": "null-result",
   "claim": "R88 Pack sources fix v2: 180 Haiku",
   "value": "+0.010 (−0.003…+0.024)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: NO as score lever (kept as fix)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L309"
   ]
  },
  {
   "id": "F-1129",
   "kind": "null-result",
   "claim": "R90 Rescue dropped health-pack source: replay",
   "value": "recovered 1/18",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: step-0 FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L311"
   ]
  },
  {
   "id": "F-1130",
   "kind": "null-result",
   "claim": "R91 Wider X2b budget on lite: 180 Haiku",
   "value": "+0.005 (−0.014…+0.025)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: NO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L312"
   ]
  },
  {
   "id": "F-1131",
   "kind": "null-result",
   "claim": "R93 X2b on the claim kit: replay",
   "value": "ceiling ≈ 0.011",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: closed at ceiling",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L314"
   ]
  },
  {
   "id": "F-1132",
   "kind": "null-result",
   "claim": "R95 4B query writer for pack search: replay",
   "value": "top-8 69 → 67",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: step-0 FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L316"
   ]
  },
  {
   "id": "F-1133",
   "kind": "null-result",
   "claim": "Raw-text Wikipedia packs for the 4B (early): vital5 leads, top-100k full",
   "value": "Δ 0.000 / −0.005",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: \"more text stopped working\"",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-results-L320"
   ]
  },
  {
   "id": "F-1134",
   "kind": "null-result",
   "claim": "Big Wikipedia pack (top-100k, 2.54 GB): as a peer pack",
   "value": "4B key-in-context +3 (bar +6), retrieval 3.3×; deep −0.034, fills 26/28 slots",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/CLOSED.md",
   "caveat": "why closed: crowding",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-system-L214"
   ]
  },
  {
   "id": "F-1135",
   "kind": "null-result",
   "claim": "Many packs at once: 9 peers",
   "value": "−0.019 Opus",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:11",
   "caveat": "why closed: crowding",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "packs",
   "aliases": [
    "N-system-L215"
   ]
  },
  {
   "id": "F-1136",
   "kind": "decision",
   "claim": "Every pass/fail criterion written and sha256-frozen before the run; no edits after verdicts, only dated addenda — why: Prevent tuning the rule to the result — outcome: 65/65 criterion hashes verified 10-10; 16 of 17 re-verified by an outside critic 10-16",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "source_public": "boar-research:method/METHOD.md:5",
   "caveat": "Verification note: \"18/19 re-verified\" → 16/17 (CRITIQUE:6) or 19/20 in the C6 spot-check (CRITIQUE:72-75).",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-process-02"
   ]
  },
  {
   "id": "F-1137",
   "kind": "decision",
   "claim": "Report a quality effect only if it exceeds the MDE printed by the diff tool — why: run-to-run noise (Δ 0.031) as large as effects — outcome: became house rule; later refined (two-seed, seed-noise rule)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "date_text": "2026-10-01 23:00",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-chron-early-04"
   ]
  },
  {
   "id": "F-1138",
   "kind": "decision",
   "claim": "Hold the cross-family (non-Claude) judge pilot — why: needs paid API + would send private references to a third party — outcome: judge-family calibration done later with a local open judge (10-07) and in a later slice",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-chron-early-13"
   ]
  },
  {
   "id": "F-1139",
   "kind": "decision",
   "claim": "Pairwise judging in both A/B orders; disagreement = tie; scores averaged across orders (why: removes position bias)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": "boar-bench:scorers/vendor/eval-lib/judge-core.mjs:93-98; boar-bench:TRACE-FORMAT.md:53",
   "caveat": "outcome: position consistency 96.5-100 % measured",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-method-05"
   ]
  },
  {
   "id": "F-1141",
   "kind": "fact",
   "claim": "Retrieval JS on the phone: pure-JS zstd decompression ~76 % of retrieval time; native zstd ~80× faster",
   "value": "499 ms of 677 ms",
   "ci": "68 calls",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": "n/a",
   "date": "2026-10-02",
   "source_public": null,
   "caveat": "led to native zstd arm (device-only gains)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-chron-early-38"
   ]
  },
  {
   "id": "F-1142",
   "kind": "fact",
   "claim": "Second Claude judge (Fable 5.1) vs Opus: same winner 453/456 verdicts, Pearson r 0.939, κ (correctness ≥ 4) 0.894, ratio +0.004…+0.010.",
   "value": "453/456",
   "ci": "456 verdicts",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Opus vs Fable",
   "date": "2026-10-02",
   "source_public": "boar-bench:TRACE-FORMAT.md:78",
   "caveat": "both judges are Claude: tests noise, not self-preference; ≈ 23 USD list",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-method-22"
   ]
  },
  {
   "id": "F-1143",
   "kind": "fact",
   "claim": "Earlier cross-family checks",
   "value": "Fable (Claude family) second judge: same winner 453/456, κ 0.894; local Gemma judge +0.11 lenient, κ 0.32/0.55",
   "ci": "n 57",
   "n": "57",
   "seeds": null,
   "set": "v0.1 (dev)",
   "set_status": "dev",
   "judge": "Fable / Gemma vs Opus",
   "date": "2026-10-02",
   "date_text": "2026-10-02, 2026-10-07",
   "source_public": null,
   "caveat": "Fable is same family",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-process-67"
   ]
  },
  {
   "id": "F-1144",
   "kind": "event",
   "claim": "Second Claude judge (Fable 5.1) agrees with Opus on 453/456 verdicts",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-method-L33"
   ]
  },
  {
   "id": "F-1145",
   "kind": "fact",
   "claim": "bmoe Android payload.",
   "value": "12.4 MB stripped (≈ 4.6 MB compressed)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": "D2/D3 (app module, engine arm) not built",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-system-48"
   ]
  },
  {
   "id": "F-1146",
   "kind": "decision",
   "claim": "Headline = floor-adjusted ratio + win/tie/loss; raw second — why: Always-decline system scores raw ≈ 0.31 for free — outcome: Outside critic: post-hoc but toward the stricter scale, \"not self-serving\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-process-08"
   ]
  },
  {
   "id": "F-1147",
   "kind": "fact",
   "claim": "Steps before the first token on a phone (cut 4 model): detector, rules, places, model fit, route gate (one LLM forward pass), query embedding, retrieval (FTS5 BM25 + brute-force cosine, no reranker), compress/ground, prompt, prefill.",
   "value": "10 steps; 2 LLM calls for 100/120 traces",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.1 traces",
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "diagram source",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-system-25"
   ]
  },
  {
   "id": "F-1148",
   "kind": "decision",
   "claim": "v0.3 written by the author with a non-Claude web model; hashes only before scoring — why: self-preference risk; burnt v0.2 — outcome: spec ready; items not written by 10-10",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-chron-early-27"
   ],
   "superseded_by": [
    "F-1374",
    "F-2070"
   ],
   "superseded_note": "No v0.3 was written (decision D5, 2026-10-18); the v0.4 ruler set (half Claude, half OpenAI gpt-5.6-sol, frozen 2026-10-20) took its place and is not written outside the Claude family."
  },
  {
   "id": "F-1149",
   "kind": "decision",
   "claim": "Haiku 5.5-low replaces Sonnet-low as L1 judge (why: same winner 54/57, 3.5 s/call, ~14× cheaper)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": "outcome: adopted; known to compress gaps",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-method-12"
   ]
  },
  {
   "id": "F-1150",
   "kind": "decision",
   "claim": "Haiku 5.5-low as L1 triage judge, Opus only to confirm and for claim numbers; judge sharded N = 4 — why: ≈ 14× cheaper than the previous triage judge; same winner as Opus 54/57 — outcome: Haiku compresses gaps 15-30 %; near-miss escalation rules added",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-process-09"
   ]
  },
  {
   "id": "F-1151",
   "kind": "decision",
   "claim": "Haiku 5.5-low as L1 triage judge; Opus only for confirmations/claim numbers (why: 14× cheaper, same ranking; compresses gaps)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": "boar-research:reports/RULE-CHANGES.md:9,14",
   "caveat": "outcome: Near-miss rule added 10-10 after Haiku stopped DEEPTHINK-V02 wrongly",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-results-05"
   ]
  },
  {
   "id": "F-1152",
   "kind": "fact",
   "claim": "Haiku 5.5-low as triage judge: same winner as Opus 54/57, κ 0.65, ~14× cheaper, compresses gaps",
   "value": "54/57; ratio offset −0.069",
   "ci": "228 calls",
   "n": null,
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Haiku vs Opus",
   "date": "2026-10-07",
   "source_public": "boar-research:reports/RULE-CHANGES.md:9",
   "caveat": "deep − 4B +0.007 under Haiku vs +0.029 Opus",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-chron-early-86"
   ]
  },
  {
   "id": "F-1153",
   "kind": "fact",
   "claim": "Haiku compresses paired gaps: deep35 − 4B was +0.007 under Haiku vs +0.029 under Opus (×0.25).",
   "value": "+0.007 vs +0.029",
   "ci": "n 57",
   "n": "57",
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "both",
   "date": "2026-10-07",
   "source_public": null,
   "caveat": "review later measured ×0.6-1.0 on full deep sets, ×3-4 on hard slices",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-method-40"
   ]
  },
  {
   "id": "F-1154",
   "kind": "event",
   "claim": "JUDGE2: local open-weights judge (Gemma 4 26B) keeps the ranking but scores +0.11 above Opus",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-method-L42"
   ]
  },
  {
   "id": "F-1155",
   "kind": "event",
   "claim": "HAIKUJUDGE: Haiku 5.5-low adopted as L1 triage judge; JUDGE2 local Gemma judge = ranking only (+0.11 lenient)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-results-L41"
   ]
  },
  {
   "id": "F-1156",
   "kind": "event",
   "claim": "HAIKUJUDGE: Haiku 5.5-low adopted as L1 judge (54/57 same winner, ~14× cheaper than Sonnet-low)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "date_text": "2026-10-07 17:58",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-method-L41"
   ]
  },
  {
   "id": "F-1157",
   "kind": "fact",
   "claim": "Recommended deep configuration (35B MoE + thinking 512) on V-para, two seeds: raw 0.553, floor-adj 0.436, win 0.",
   "value": "0.553 / 0.436",
   "ci": "raw 0.524-0.581; floor 0.401-0.472; common-mode floor 0.387-0.486; n 51",
   "n": "51",
   "seeds": null,
   "set": "V-para, dev-style",
   "set_status": "dev-style",
   "judge": "Opus",
   "date": "2026-10-09",
   "source_public": "boar-research:reports/BAR-TABLE.md:39,126",
   "caveat": "citation fix: H row l.39, common-mode CI l.126 (cited l.37 = D row, l.141 = pipeline metric); winner's curse ≈ 0.01-0.02",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-method-28"
   ]
  },
  {
   "id": "F-1158",
   "kind": "decision",
   "claim": "Near-miss rule: a Haiku Δ within 0.01 of a threshold triggers Opus on the same ids — why: Haiku compressed a +0.048 Opus gap to +0.027 — outcome: applied to THINK4B",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": "boar-research:reports/RULE-CHANGES.md:14",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-chron-early-32"
   ]
  },
  {
   "id": "F-1159",
   "kind": "decision",
   "claim": "Seed-noise reading rule (V-para one-seed deep Δ < ~0.03 is noise; later raised to 0.05) — why: SEEDVAR three-seed spread — outcome: in CLAUDE.md and RULE-CHANGES",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": "boar-research:reports/RULE-CHANGES.md:16",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-chron-early-33"
   ]
  },
  {
   "id": "F-1160",
   "kind": "decision",
   "claim": "Seed-noise reading rule: one-seed deep Δ on V-para < 0.03 = noise; claim numbers use ≥ 2 seeds (why: SEEDVAR SD 0.020)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": "boar-research:method/L1-PROTOCOL.md:27",
   "caveat": "outcome: raised to ≥ 0.05 on 2026-10-16",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-method-18"
   ]
  },
  {
   "id": "F-1161",
   "kind": "decision",
   "claim": "Random stratified 20-item slices; futility = stop only if Haiku Δ ≤ 0 and head-to-head ≤ control — why: \"Hardest 20\" slices inflated effects (+0.091 → +0.026; +0.089 → +0.009) — outcome: Later critic: one n = 20 Haiku pass carries ±0.03-0.04 noise, about the MDE",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-process-13"
   ]
  },
  {
   "id": "F-1162",
   "kind": "decision",
   "claim": "Seed-noise rule: a one-seed V-para deep Δ < 0.03 (raised to < 0.05 on 10-16) is seed noise (why: SEEDVAR SD 0.020 raw / 0.025 floor-adj; seeds 42–43 differ −0.039 with CI excluding 0)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": "boar-research:reports/RULE-CHANGES.md:16",
   "caveat": "outcome: Two-seed V-para confirmation required for adoption",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-results-08"
   ]
  },
  {
   "id": "F-1163",
   "kind": "event",
   "claim": "SEEDVAR: three-seed spread of the deep tier; seed-noise rule",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "E-chron-early-L112"
   ]
  },
  {
   "id": "F-1164",
   "kind": "fact",
   "claim": "Best-of-2 oracle gain (+0.038) is judge noise + seed gap; no judge-blind selector captured it",
   "value": "self-selection agreement 33 %, oracle agreement 49 %",
   "ci": "51 items, 2 seeds",
   "n": null,
   "seeds": "2",
   "set": "V-para dev-style",
   "set_status": "dev-style",
   "judge": "existing Opus rows",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/CLOSED.md:46",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-chron-early-101"
   ]
  },
  {
   "id": "F-1165",
   "kind": "fact",
   "claim": "Thinking twin of the 4B: small, inside seed noise",
   "value": "V-para Δ +0.026 (−0.013…+0.068); v0.2 +0.009 (−0.015…+0.035)",
   "ci": "1 seed each",
   "n": null,
   "seeds": "1",
   "set": "V-para dev-style (Opus); v0.2 dev (Haiku)",
   "set_status": "dev-style",
   "judge": "Opus / Haiku L1",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/CLOSED.md:25-26",
   "caveat": "×3.7 time",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-chron-early-102"
   ]
  },
  {
   "id": "F-1166",
   "kind": "fact",
   "claim": "On the 20 hardest items Haiku showed +0.029 where Opus gave +0.091; on the full set they agreed within 0.005 (+0.022 vs +0.026).",
   "value": "×3 compression on slice",
   "ci": "n 20 / 51",
   "n": "20",
   "seeds": null,
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "both",
   "date": "2026-10-10",
   "source_public": "boar-research:method/L1-PROTOCOL.md:26",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-method-43"
   ]
  },
  {
   "id": "F-1167",
   "kind": "fact",
   "claim": "Two-seed deep numbers carry ≈ ± 0.014 (V-para) and ± 0.01 (v0.2) from seed noise.",
   "value": "± 0.014 / ± 0.01",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-10",
   "source_public": "boar-research:method/L1-PROTOCOL.md:27",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-method-96"
   ]
  },
  {
   "id": "F-1168",
   "kind": "fact",
   "claim": "Seed noise is large on the deep tier",
   "value": "SD 0.020 raw / 0.025 floor-adj; seed-to-seed paired Δ −0.039 (−0.072…−0.011)",
   "ci": "3 seeds, n 51",
   "n": "51",
   "seeds": "3",
   "set": "V-para (dev-style)",
   "set_status": "dev-style",
   "judge": "Opus",
   "date": "2026-10-10",
   "date_text": "2026-10-10/16",
   "source_public": "boar-research:reports/RULE-CHANGES.md:16",
   "caveat": "rule raised from 0.03 to 0.05 after critique",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-process-31"
   ]
  },
  {
   "id": "F-1169",
   "kind": "fact",
   "claim": "Criterion hashes verified: 65/65 on 2026-10-10; an outside critic found 16 of 17 matching on 2026-10-16.",
   "value": "65/65 (10-10); outside critic: 16/17 criterion hashes match (verdict), 19 of 20 in the C6 spot-check (1 mismatch by hashing convention, 1 without a .sha256 sidecar)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10, 2026-10-16",
   "source_public": null,
   "caveat": "some early freezes attested by mtime only",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-process-55"
   ]
  },
  {
   "id": "F-1170",
   "kind": "fact",
   "claim": "Thinking 4B on v0.2",
   "value": "Δ +0.009",
   "ci": "−0.015…+0.035; 1 seed",
   "n": "110 (inferred from set)",
   "seeds": "1",
   "set": "v0.2 dev",
   "set_status": "dev",
   "judge": "Haiku L1",
   "date": "2026-10-10",
   "source_public": "boar-research:reports/CLOSED.md:26",
   "caveat": "Haiku compresses; Opus never run",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-results-35"
   ]
  },
  {
   "id": "F-1171",
   "kind": "event",
   "claim": "SEEDVAR: H over 3 seeds mean floor-adj 0.432 (SD 0.025); one-seed deep Δ < 0.03 = seed noise",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-results-L52"
   ]
  },
  {
   "id": "F-1172",
   "kind": "event",
   "claim": "BAR-TABLE written (all configs vs bar); LENGTHSHARE: ≈ 17 % of H's gap is reference length, ≈ 83 % substance",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "source_public": "boar-research:reports/BAR-TABLE.md:1-98",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-results-L53"
   ]
  },
  {
   "id": "F-1173",
   "kind": "event",
   "claim": "SEEDVAR: three seeds of the deep configuration on V-para: SD 0.020 raw / 0.025 floor-adj",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 12:33",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-method-L50"
   ]
  },
  {
   "id": "F-1174",
   "kind": "event",
   "claim": "R&D process review (10 Opus agents: 7 readers, 2 refuters, 1 critic; 76 findings): \"measurement machine better than typical, pointed at the wrong target\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 13:30",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "E-process-L44"
   ]
  },
  {
   "id": "F-1175",
   "kind": "event",
   "claim": "LENGTHSHARE: re-judging against 80-word references; ≈17 % of the gap is reference length, ≈83 % missing substance; first non-zero win-rate (3/51) only against shortened refs",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 22:31–22:36",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "E-chron-late-L26"
   ]
  },
  {
   "id": "F-1176",
   "kind": "decision",
   "claim": "Common-mode CIs for deep-tier rows: CI = Δ ± 1.96·sqrt(SE_item² + 2σ_run²) (why: item bootstrap ignores run-level seed noise)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "source_public": "boar-research:reports/RULE-CHANGES.md:23",
   "caveat": "outcome: deep CIs ~1.2-1.3× wider; readings unchanged",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-method-23"
   ]
  },
  {
   "id": "F-1177",
   "kind": "decision",
   "claim": "Relabel under-powered negatives \"not detected at MDE x (n, seeds)\"; keep a closed-registry with what reopens each family — why: Closed ≠ disproven — outcome: Public registry R1-R76; critic still found 7 rows to relabel",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "source_public": "boar-research:reports/CLOSED.md:6-9",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-process-17"
   ]
  },
  {
   "id": "F-1178",
   "kind": "decision",
   "claim": "Relabel under-powered negatives \"not detected at MDE x (n, seeds)\" instead of \"closed\" (why: honesty review: several closures could not exclude a +0.03 lever)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "source_public": "boar-research:reports/RULE-CHANGES.md:25",
   "caveat": "the critique names 7 n≈20 closures with CI upper ≥ +0.03 (R10, R25, R31, R83, R84, R85, R97); R10 and R25 were already 'not detected', so 5 still need the relabel (cited l.47-50 are about exposure)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-results-09"
   ]
  },
  {
   "id": "F-1179",
   "kind": "fact",
   "claim": "Judge test-retest on unchanged answers: raw 0.533 → 0.537 (+0.004, −0.005…+0.013); winner agreement 51/51; per-item r 0.96.",
   "value": "σ_judge ≈ 0.003 raw / 0.004 floor",
   "ci": "n 51, 102 Opus",
   "n": "51,",
   "seeds": null,
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-11",
   "source_public": "boar-research:reports/BAR-TABLE.md:123",
   "caveat": "public citation fix: l.136 is a separator",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-method-37"
   ]
  },
  {
   "id": "F-1180",
   "kind": "event",
   "claim": "JUDGEVAR: Opus test-retest σ ≈ 0.003 (negligible); common-mode CI rule adopted",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "source_public": "boar-research:reports/BAR-TABLE.md:122-131",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-results-L55"
   ]
  },
  {
   "id": "F-1181",
   "kind": "event",
   "claim": "JUDGEVAR: Opus judge test-retest σ ≈ 0.003 raw; common-mode CI flag shipped",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "date_text": "2026-10-11 03:00",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "E-chron-late-L29"
   ]
  },
  {
   "id": "F-1182",
   "kind": "fact",
   "claim": "Opus judge test-retest noise is tiny compared with seed noise (JUDGEVAR)",
   "value": "σ_judge ≈ 0.003 raw / 0.004 floor-adj vs σ_run 0.020 / 0.025",
   "ci": "per-item r 0.96; 102 calls",
   "n": null,
   "seeds": null,
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "Opus",
   "date": "2026-10-11",
   "date_text": "2026-10-11 03:00",
   "source_public": "boar-research:reports/BAR-TABLE.md:122",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-chron-late-08"
   ]
  },
  {
   "id": "F-1183",
   "kind": "event",
   "claim": "JUDGEVAR: re-judging unchanged answers gives σ_judge ≈ 0.003; common-mode CIs adopted for deep-tier rows",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "date_text": "2026-10-11 03:00",
   "source_public": "boar-research:reports/RULE-CHANGES.md:23",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-method-L54"
   ]
  },
  {
   "id": "F-1184",
   "kind": "fact",
   "claim": "V-para remaining loss composition (deep tier)",
   "value": "59 % absent knowledge + 40 % wrong facts",
   "ci": null,
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para dev",
   "set_status": "dev",
   "judge": "jev labels (LOSSMAP)",
   "date": "2026-10-12",
   "source_public": "boar-research:reports/BAR-TABLE.md:142",
   "caveat": "label model unvalidated (critique C3)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-results-44"
   ]
  },
  {
   "id": "F-1185",
   "kind": "decision",
   "claim": "Never idle while 0-paid-call work exists; no Haiku cap — why: Author objected after 9 idle ticks (\"why idle when I asked to never stop?\", paraphrase) — outcome: 10-14 log: 0 idle ticks; later critic warns \"idle lane = failure\" rewards trivial probes",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-process-18"
   ]
  },
  {
   "id": "F-1186",
   "kind": "fact",
   "claim": "Haiku and Opus agreed on the winner almost always on dev runs",
   "value": "359/360 and 179/180 items",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "180 self-tests (dev)",
   "set_status": "dev",
   "judge": "Haiku vs Opus",
   "date": "2026-10-13",
   "date_text": "2026-10-13/14",
   "source_public": null,
   "caveat": "critic: \"same winner\" carries no information because the reference wins ≈ 100 %",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-process-27"
   ]
  },
  {
   "id": "F-1187",
   "kind": "fact",
   "claim": "Haiku L1 compresses paired gaps vs Opus",
   "value": "Haiku Δ ≈ 0.70-0.95× the Opus Δ across six pack/retrieval arms (X1 0.70, pack 0.76, ONLY 0.83, X2b 0.85, ST2 0.88, ADD 0.95); per-item Δ r 0.75-0.84",
   "ci": "several arms",
   "n": "180 (inferred from set)",
   "seeds": null,
   "set": "180 self-tests",
   "set_status": null,
   "judge": "Haiku vs Opus",
   "date": "2026-10-14",
   "date_text": "2026-10-14/16",
   "source_public": null,
   "caveat": "citation fix (critique row is l.32); '0.70-0.85' are the two quoted examples; on deep/4B slices compression ranged ×0.25 to a sign flip",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-results-55"
   ]
  },
  {
   "id": "F-1188",
   "kind": "decision",
   "claim": "Apply 7 re-wordings from the adversarial critique (retract \"gap is reach, not reading\", etc.) — why: Numbers real, summaries scope-stripped — outcome: Corrections logged as QUEUE rows",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-process-26"
   ]
  },
  {
   "id": "F-1189",
   "kind": "decision",
   "claim": "Name the judge family next to every ratio; lead with the stricter (Claude) judge; keep win-rate first — why: Cross-family judge ≈ +0.10 more lenient — outcome: Claim v4.2.x",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16/17",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-process-27"
   ]
  },
  {
   "id": "F-1190",
   "kind": "decision",
   "claim": "Lite tier proposed for 4–6 GB phones: Qwen3-1.7B behind the full safety engine (why: 1.1 GB file, RSS ≈ 2.3 GB; safety routing identical to 4B)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "outcome: offered as option; quality wording corrected after critique",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-system-19"
   ],
   "superseded_by": [
    "F-1386",
    "F-1349"
   ],
   "superseded_note": "Superseded on 2026-10-20: Qwen3-1.7B became the main 8 GB engine (not only a lite tier for 4-6 GB phones); the 4B became the optional tier."
  },
  {
   "id": "F-1191",
   "kind": "event",
   "claim": "Adversarial outside critique: numbers real, internal narrative overstated; 7 re-wordings applied 18:28",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 evening",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "E-process-L71"
   ]
  },
  {
   "id": "F-1192",
   "kind": "fact",
   "claim": "Haiku compresses effect sizes",
   "value": "Haiku Δ = 0.70× and 0.85× the Opus Δ; per-item r 0.75-0.84; level −0.069",
   "ci": "2 arms",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Haiku vs Opus",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-process-28"
   ]
  },
  {
   "id": "F-1193",
   "kind": "fact",
   "claim": "Haiku n = 20 judge noise is about the size of the decision threshold",
   "value": "±0.03-0.04 per pass; per-item retest SD 0.22-0.28 rubric points",
   "ci": "n 18-30 identical answers",
   "n": "18",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Haiku",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-process-32"
   ]
  },
  {
   "id": "F-1194",
   "kind": "fact",
   "claim": "Outside critic re-derived numbers",
   "value": "15/17 traced numbers exact or in the cited file; 2 \"different\" (stale/scope-stripped); no invented numbers",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "V-para/v0.2 rows checked on aggregates only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-process-57"
   ]
  },
  {
   "id": "F-1195",
   "kind": "fact",
   "claim": "A single n=20 Haiku pass carries ±0.03–0.04 noise",
   "value": "per-item retest SD 0.22–0.28 rubric points",
   "ci": "n 18–30",
   "n": "18",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Haiku",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "citation fix (l.34 is the self-preference row) | at the MDE: n=20 labels are noise-level",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-results-56",
    "F-chron-late-88"
   ]
  },
  {
   "id": "F-1196",
   "kind": "event",
   "claim": "Critique raises the one-seed noise rule from 0.03 to ≥ 0.05",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-method-L63"
   ]
  },
  {
   "id": "F-1197",
   "kind": "event",
   "claim": "Adversarial critique: lite \"= 4B\" overstated (kit −0.111), lite gap is reading too (−0.070 at identical contexts); closed loop risk",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-results-L71"
   ]
  },
  {
   "id": "F-1198",
   "kind": "event",
   "claim": "Lite tier (Qwen3-1.7B) evaluated; LITESAFETY PASS; critique corrects \"1.7B = 4B\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-system-L64"
   ]
  },
  {
   "id": "F-1199",
   "kind": "decision",
   "claim": "Idle-lane rule: a blocked lane is one log row; every other lane filled at once; refined 21:5x to \"arms only for new questions with ceiling ≥ MDE, quiet ticks otherwise\" — why: ≈ 5 h of idle lanes on 10-16 — outcome: settled the critique's point on 3-minute filler probes",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 09:4x",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-chron-late-35"
   ]
  },
  {
   "id": "F-1200",
   "kind": "fact",
   "claim": "Small-model router ceilings (SMALLROUTE S0)",
   "value": "lite ceiling 0.018 on the 180; deep ceiling 0.110 on v0.2; no 0-call signal beats chance",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "the 180 + v0.2 dev",
   "set_status": "dev",
   "judge": "Opus/Haiku rows reused",
   "date": "2026-10-16",
   "date_text": "2026-10-16 18:07",
   "source_public": null,
   "caveat": "ceilings in-sample, inflated by judge noise",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-chron-late-96"
   ]
  },
  {
   "id": "F-1201",
   "kind": "decision",
   "claim": "Apply the adversarial critique: lite \"= 4B\" only on pack-matched dev; one-seed V-para noise rule 0.03 → 0.05; n≈20 closures relabelled \"not detected at MDE 0.08\" — why: over-stated internal summaries — outcome: 7 re-wordings applied",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 18:28",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "D-chron-late-37"
   ]
  },
  {
   "id": "F-1202",
   "kind": "event",
   "claim": "XFAM: cross-family judge (OpenAI gpt-5.6-sol via the Codex CLI, 408 calls) on V-para traces: ≈ +0.10 vs Claude, same ranking, win-rate ≈ 0",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 19:01",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-method-L64"
   ]
  },
  {
   "id": "F-1203",
   "kind": "event",
   "claim": "XFAM: cross-family (OpenAI) judge scores both arms ≈ +0.10 higher on V-para; deep floor-adj 0.577",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 19:01",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-results-L72"
   ]
  },
  {
   "id": "F-1204",
   "kind": "event",
   "claim": "XFAM: cross-family judge on V-para scores both arms ≈ +0.10 higher",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 19:02",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "E-chron-late-L95"
   ]
  },
  {
   "id": "F-1205",
   "kind": "event",
   "claim": "Claim v4.2 / v4.2.1: judge family named on every ratio; native comparison kept internal",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 21:00–21:44",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "E-chron-late-L96"
   ]
  },
  {
   "id": "F-1206",
   "kind": "fact",
   "claim": "Win-rate vs the reference under the cross-family judge on v0.2: 4B 2.7 %, deep 6.4 %; under Claude 0 %.",
   "value": "2.7 % / 6.4 % / 0 %",
   "ci": "n 110",
   "n": "110",
   "seeds": null,
   "set": "v0.2 dev",
   "set_status": "dev",
   "judge": "both",
   "date": "2026-10-17",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-method-56"
   ]
  },
  {
   "id": "F-1207",
   "kind": "fact",
   "claim": "Self-corrections recorded in the log",
   "value": "43 rows mention a correction (critic counted 29 CORRECTION rows)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "source_public": null,
   "caveat": "count method differs",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-process-58"
   ]
  },
  {
   "id": "F-1208",
   "kind": "fact",
   "claim": "Win-rate under the cross-family judge on v0.2",
   "value": "4B 2.7 %, deep 6.4 % (Claude 0 %)",
   "ci": "n 110",
   "n": "110",
   "seeds": null,
   "set": "v0.2 dev",
   "set_status": "dev",
   "judge": "cross-family",
   "date": "2026-10-17",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-results-52"
   ]
  },
  {
   "id": "F-1209",
   "kind": "event",
   "claim": "XFAM-V02: offset holds on v0.2",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "date_text": "2026-10-17 01:09",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "E-chron-late-L97"
   ]
  },
  {
   "id": "F-1210",
   "kind": "event",
   "claim": "XFAM-V02: same judge on v0.2 (660 calls): offset holds (+0.10 raw / +0.13 floor-adj), deep − 4B gap identical",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "date_text": "2026-10-17 01:09",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-method-L65"
   ]
  },
  {
   "id": "F-1211",
   "kind": "event",
   "claim": "XFAM-V02: same offset on v0.2; deep floor-adj 0.639 under OpenAI judge; ranking unchanged",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "date_text": "2026-10-17 01:09",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-results-L74"
   ]
  },
  {
   "id": "F-1212",
   "kind": "event",
   "claim": "Claim v4.2.3: \"the lower judge is the headline\"; judge family named next to every ratio",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "date_text": "2026-10-17 20:52",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "T-method-L66"
   ]
  },
  {
   "id": "F-1213",
   "kind": "fact",
   "claim": "74 of 102 re-judged calls were identical on all three axes; correctness moves most (mean",
   "value": "Δ",
   "ci": "0.147).",
   "n": null,
   "seeds": null,
   "set": "74/102",
   "set_status": null,
   "judge": null,
   "date": null,
   "date_text": "V-para dev",
   "source_public": null,
   "caveat": "coord/arms/judgevar/JUDGEVAR.md:3,9",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "F-method-38"
   ]
  },
  {
   "id": "F-1214",
   "kind": "null-result",
   "claim": "Thinking 4B (THINK4B, -V02): Qwen3-4B-Thinking at 256/512 → +0.026 / +0.009 (below +0.03; seed noise)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "N-chron-early-L330"
   ]
  },
  {
   "id": "F-1215",
   "kind": "null-result",
   "claim": "Haiku medium effort: second leg of HAIKUJUDGE",
   "value": "not run",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: low effort already passed the agreement bar",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "N-method-L214"
   ]
  },
  {
   "id": "F-1216",
   "kind": "null-result",
   "claim": "Changing the judge to another Claude model: Fable second judge",
   "value": "moves the ratio < 0.01",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "citation fix (l.27 is a table header); why: does not address self-preference",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "N-method-L223"
   ]
  },
  {
   "id": "F-1217",
   "kind": "null-result",
   "claim": "\"Idle lane = failure\" as a target: Utilisation metric → critic: rewards trivial 3-minute probes (Recommend information per judge call instead)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "N-process-L208"
   ]
  },
  {
   "id": "F-1218",
   "kind": "null-result",
   "claim": "Cross-family judge via paid API (GPT pilot): 204-call pilot → held / withdrawn (no paid keys, no human review) (Later done by the author via Codex (XFAM))",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "measurer",
   "aliases": [
    "N-process-L212"
   ]
  },
  {
   "id": "F-1219",
   "kind": "decision",
   "claim": "No human review: an independent Opus reviewer agent decides content and wording — why: Author's time; consistency — outcome: Every claim draft, page and release passed reviewer rounds; \"AI-reviewed, not clinically reviewed\" caveat kept",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "D-process-06"
   ]
  },
  {
   "id": "F-1220",
   "kind": "decision",
   "claim": "Adopt cut 3 (verdict guard + arithmetic repair + snakebite text) (why: L1 Δ +0.029 [+0.007, +0.058], confident-wrong 15 → 9, unsafe 0)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "source_public": null,
   "caveat": "outcome: Cut 3 became the first claim build; claim runs 10-05",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "D-results-01"
   ]
  },
  {
   "id": "F-1221",
   "kind": "decision",
   "claim": "Adopt cut 3 (verdict guard v3 + arithmetic tools + snakebite text) (why: L1 gain +0.029 and reviewer PASS on fresh probes)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-04",
   "source_public": null,
   "caveat": "outcome: cut 3 used for first claim runs and first phone",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "D-system-03"
   ]
  },
  {
   "id": "F-1222",
   "kind": "event",
   "claim": "First claim runs (cut 3): held-out v0.2 raw 0.513; V-para raw 0.413",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "E-chron-early-L86"
   ]
  },
  {
   "id": "F-1223",
   "kind": "fact",
   "claim": "Haiku without web scores below the floor-adjusted bar too",
   "value": "raw 0.588 (0.559-0.615) / floor-adj 0.48",
   "ci": "n 51",
   "n": "51",
   "seeds": null,
   "set": "V-para (dev-style)",
   "set_status": "dev-style",
   "judge": "Claude Opus",
   "date": "2026-10-05",
   "source_public": "boar-research:README.md:30",
   "caveat": "shows the internal bar is stricter than a small frontier model offline | Verification note: The Haiku no-web anchor was run with the cut 3 claim runs on 10-05 (QUEUE:642), not 10-06. One seed; floor-adj CI unknown.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "F-process-08"
   ]
  },
  {
   "id": "F-1224",
   "kind": "fact",
   "claim": "Cut 3 claim run on v0.2: raw",
   "value": "0.513",
   "ci": "0.479–0.550; n 110; 2 seeds",
   "n": "110",
   "seeds": "2",
   "set": "v0.2 held-out at time",
   "set_status": "held-out at run time",
   "judge": "Opus",
   "date": "2026-10-05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "F-results-08"
   ]
  },
  {
   "id": "F-1225",
   "kind": "event",
   "claim": "4B claim runs on V-para (cut 3, cut 6) while V-para was still unseen: raw 0.413, floor-adj 0.261, win 0/51",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-05",
   "date_text": "2026-10-05 / 10-06",
   "source_public": "boar-research:reports/BAR-TABLE.md:35; boar-research:reports/SET-STATUS.md:15",
   "caveat": "citation fix only: BAR-TABLE l.31 is the section header, the K row is l.35",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "T-method-L38"
   ]
  },
  {
   "id": "F-1226",
   "kind": "decision",
   "claim": "Headline = floor-adjusted ratio + win/tie/loss (+ share of items ≥ 3/5), raw second; own bar = floor-adj ≥ 0.50 on V-para — why: raw 0.50 is ~61 % free because of the rubric floor — outcome: board and claim draft rewritten 10-06 22:40",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "D-chron-early-21"
   ]
  },
  {
   "id": "F-1227",
   "kind": "decision",
   "claim": "Headline = floor-adjusted ratio + win/tie/loss (+ share useful ≥ 3/5); raw second (why: raw 0.50 is ~61 % \"free\" (always-decline ≈ 0.306))",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": "citation fix (D1 is l.53). Date nuance: decided 10-06 in the audit, applied to the claim draft v3.1 on 10-07 (CRITIQUE-2026-10-16.md:24); floor scale pre-registered as companion 10-02 (analysis-plan.md:22-25)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "D-method-09"
   ]
  },
  {
   "id": "F-1228",
   "kind": "event",
   "claim": "Cut 6 claim runs (second spend): v0.2 0.512 (0.478–0.549), V-para 0.413 (0.384–0.445), floor-adj 0.261, Haiku-no-web anchor 0.588",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "T-results-L34"
   ]
  },
  {
   "id": "F-1229",
   "kind": "event",
   "claim": "Cut 6 claim runs: v0.2 raw 0.512, V-para raw 0.413 / floor-adj 0.261",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": "boar-research:README.md:25-26",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "T-system-L45"
   ]
  },
  {
   "id": "F-1230",
   "kind": "decision",
   "claim": "Freeze cut 7 = cut 6 + FACTS r6 on deterministic gates alone; full claim runs deferred to v0.3 (why: FACTS fires 0/110 and 0/58 on the claim sets, so claim numbers are cut 6's)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": "outcome: Cut 7 is the frozen claim build ever since",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "D-results-02"
   ]
  },
  {
   "id": "F-1231",
   "kind": "fact",
   "claim": "Desktop latency per answer (p50) along the ladder on V-para",
   "value": "4B 6.3 s (same backend) / 10.1 s (claim run); D 33.3/30.9 s; Q 57.5/53.2 s; H 78.9/79.4 s; G 106–109 s; T 142.9/146.0 s; Z 23.6 s",
   "ci": "per seed",
   "n": "51 (inferred from set)",
   "seeds": null,
   "set": "V-para",
   "set_status": null,
   "judge": null,
   "date": "2026-10-08",
   "date_text": "2026-10-08/10",
   "source_public": "boar-research:reports/BAR-TABLE.md:35-43",
   "caveat": "Mac mini M4 CPU desktop proxy, not a phone",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "F-results-30"
   ]
  },
  {
   "id": "F-1232",
   "kind": "decision",
   "claim": "V-para status: held-out for the 4B claim runs; dev-style for everything measured after 10-06 — why: exposure ledger showed conflicting status labels — outcome: every deep-tier number on V-para is labelled dev-style",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "source_public": "boar-research:reports/SET-STATUS.md:15",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "D-chron-early-30"
   ]
  },
  {
   "id": "F-1233",
   "kind": "decision",
   "claim": "V-para = held-out for the 4B claim runs only; dev-style for all later work (why: deep-tier arms were tuned on it)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "source_public": "boar-research:reports/SET-STATUS.md:15",
   "caveat": "outcome: every deep-tier V-para number labelled development evidence",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "D-method-16"
   ]
  },
  {
   "id": "F-1234",
   "kind": "event",
   "claim": "V-para status ruling: held-out for the 4B claim runs, dev-style for everything measured after 2026-10-06",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-09",
   "source_public": "boar-research:reports/SET-STATUS.md:15",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "T-method-L47"
   ]
  },
  {
   "id": "F-1235",
   "kind": "event",
   "claim": "Handoff for the next session written: 10-step \"claim path\" (backups, decision note, UI wiring, judge fail-closed, APKs, repro package, claim text, readiness gate)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 20:57",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "E-chron-late-L18"
   ]
  },
  {
   "id": "F-1236",
   "kind": "decision",
   "claim": "Common-mode CI for deep-tier rows (adds run-level seed noise) (why: σ_run dominates σ_judge (0.020 vs 0.003))",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-11",
   "source_public": "boar-research:reports/BAR-TABLE.md:122-131",
   "caveat": "outcome: H V-para raw CI widens to 0.513–0.592; formal raw rule still met by 0.013",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "D-results-10"
   ]
  },
  {
   "id": "F-1237",
   "kind": "decision",
   "claim": "Keep the native-app comparison internal until the final build is scored (why: intermediate build, decision by the author)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15/16",
   "source_public": null,
   "caveat": "outcome: Number withheld from claim text",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "D-results-18"
   ]
  },
  {
   "id": "F-1238",
   "kind": "decision",
   "claim": "Accept the maintainer's native Kotlin app as the official submission; our numbers are labelled as the TypeScript engine's — why: the maintainer owns the submission — outcome: claim v4.1 vehicle paragraph; native score pending",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 11:00",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "D-chron-late-30"
   ],
   "superseded_by": [
    "F-1703",
    "F-1686"
   ],
   "superseded_note": "The submission changed on 2026-10-24: BoarLite, a new closed app by the author, replaces the app maintainer's native app as the bounty submission."
  },
  {
   "id": "F-1239",
   "kind": "decision",
   "claim": "Claude judge stays the headline (lower of two judges); judge family named next to every ratio; win/tie/loss leads — why: analysis-plan rule: when judges differ by > 0.05, the lower is the headline — outcome: claim v4.2.x",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16/17",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "D-chron-late-39"
   ]
  },
  {
   "id": "F-1240",
   "kind": "decision",
   "claim": "Headline judge = Claude Opus (the stricter, lower judge); cross-family numbers reported beside, labelled post-hoc (why: analysis plan's lower-judge rule)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16/17",
   "source_public": null,
   "caveat": "outcome: Claim draft v4.2.x",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "D-results-17"
   ]
  },
  {
   "id": "F-1241",
   "kind": "fact",
   "claim": "Same judge, deep config on V-para: raw 0.667 (0.621-0.711) / floor-adj 0.577 (0.519-0.632) vs Claude 0.553 / 0.436.",
   "value": "+0.115 raw / +0.141 floor",
   "ci": "n 51, 2 seeds",
   "n": "51,",
   "seeds": "2",
   "set": "V-para dev-style",
   "set_status": "dev-style",
   "judge": "cross-family",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "crosses 0.50 floor-adj only under the cross-family judge; lower-judge rule applies",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "F-method-52"
   ]
  },
  {
   "id": "F-1243",
   "kind": "decision",
   "claim": "When judges from two families disagree, the lower number is the headline; judge family named next to every ratio (why: XFAM offset ≈ +0.10)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "source_public": "boar-bench:bench/v0.3/plan/analysis-plan.md:43",
   "caveat": "outcome: claim v4.2.3",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "D-method-27"
   ]
  },
  {
   "id": "F-1244",
   "kind": "decision",
   "claim": "Claim text: native app ports retrieval and prompts, not the health path or deterministic answers (why: maintainer's own portal report + MAINT-V01 missing guard texts)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "source_public": null,
   "caveat": "outcome: claim v4.2.4",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "D-system-22"
   ]
  },
  {
   "id": "F-1245",
   "kind": "fact",
   "claim": "Claim text sentence (draft v4.2.4, as of 2026-10-17)",
   "value": "\"BOAR is not shown to be half as good as frontier AI with web.\"",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "source_public": null,
   "caveat": "draft, not published at time of reading",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "F-chron-late-84"
   ]
  },
  {
   "id": "F-1246",
   "kind": "fact",
   "claim": "The desktop runner executed the real engine on 54 questions in 2 min 34 s, 0 errors",
   "value": "2:34",
   "ci": "54 items",
   "n": null,
   "seeds": null,
   "set": "58-item bounty-style dev set",
   "set_status": "dev",
   "judge": "n/a",
   "date": "2026-10-01",
   "source_public": null,
   "caveat": "native parts shimmed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "F-chron-early-09"
   ]
  },
  {
   "id": "F-1247",
   "kind": "event",
   "claim": "No more work inside the app maintainer's app; build \"BOAR Bench\" measurement app",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-03",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "E-chron-early-L71"
   ]
  },
  {
   "id": "F-1248",
   "kind": "decision",
   "claim": "Treat v0.2 as burnt (dev); a final number needs v0.3 written outside the Claude family — why: plaintext in private git history + 4 runs + a targeted fix — outcome: v0.3 spec written 10-07; items not written by 10-10",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "D-chron-early-22"
   ],
   "superseded_by": [
    "F-1374",
    "F-2070"
   ],
   "superseded_note": "Partly superseded: v0.2 stays burnt (dev); the \"needs a v0.3\" path was dropped on 2026-10-18 and replaced by the v0.4 ruler set."
  },
  {
   "id": "F-1249",
   "kind": "decision",
   "claim": "Treat v0.2 as burnt → dev; final number needs a v0.3 written outside the Claude family (why: plaintext in git history, 4 runs, Claude-written items)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": "boar-research:reports/SET-STATUS.md:14",
   "caveat": "citation fix: D5 is l.57",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "D-method-10"
   ],
   "superseded_by": [
    "F-1374",
    "F-2070"
   ],
   "superseded_note": "Partly superseded: v0.2 stays burnt (dev); the \"needs a v0.3\" path was dropped on 2026-10-18 and replaced by the v0.4 ruler set."
  },
  {
   "id": "F-1250",
   "kind": "decision",
   "claim": "Declare v0.2 burnt → dev after its plaintext was found in git history — why: Honesty: builders could have seen it — outcome: An unexposed held-out number needs a v0.3 written outside the Claude family (0 items as of 10-17; open front OF-3)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-06",
   "source_public": "boar-research:reports/SET-STATUS.md:14",
   "caveat": "Verification note: Public v0.2 row is SET-STATUS:14 (not :17).",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "D-process-07"
   ],
   "superseded_by": [
    "F-1374",
    "F-2070"
   ],
   "superseded_note": "Partly superseded: v0.2 stays burnt (dev); the \"needs a v0.3\" path was dropped on 2026-10-18 and replaced by the v0.4 ruler set."
  },
  {
   "id": "F-1251",
   "kind": "decision",
   "claim": "v0.3 = author writes items by hand with one non-Claude web-UI model; references from the same non-Claude model; Claude never edits item text (why: removes judge-family authorship of items and references)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": "outcome: write kit delivered 10-16; no items as of 10-17",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "D-method-15"
   ],
   "superseded_by": [
    "F-1374",
    "F-2070"
   ],
   "superseded_note": "No v0.3 was written (decision D5, 2026-10-18); the v0.4 ruler set (half Claude, half OpenAI gpt-5.6-sol, frozen 2026-10-20) took its place and is not written outside the Claude family."
  },
  {
   "id": "F-1252",
   "kind": "fact",
   "claim": "v0.3 target composition: 110 items in the same 10 strata, ≥ 40 % centred outside Europe/North America, ≤ 2 per country, each hinging on a checkable fact.",
   "value": "110",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.3 (unwritten)",
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": "boar-bench:bench/v0.3/DATASET_CARD.md:14; boar-bench:CONTRIBUTING.md:29",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "F-method-86"
   ]
  },
  {
   "id": "F-1253",
   "kind": "fact",
   "claim": "v0.3 references from a non-Claude family move the denominator: v0.3 ratios are not 1:1 comparable with v0.1/v0.2.",
   "value": "not comparable",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.3",
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": "boar-bench:bench/v0.3/DATASET_CARD.md:37",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "F-method-88"
   ]
  },
  {
   "id": "F-1254",
   "kind": "event",
   "claim": "Bench v0.3 spec accepted: claim set to be written outside the Claude family by the author with a non-Claude web-UI model",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "T-method-L44"
   ]
  },
  {
   "id": "F-1255",
   "kind": "event",
   "claim": "Claim readiness audit: evidence complete except v0.3; submission side not started",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-10",
   "date_text": "2026-10-10 17:34",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "E-chron-early-L114"
   ]
  },
  {
   "id": "F-1256",
   "kind": "event",
   "claim": "v0.3 power statement: bar test at n = 110 needs a point estimate ≈ 0.53 (≈ 0.545 with one-seed run noise)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "source_public": "boar-bench:bench/v0.3/plan/power-statement.md:1-19",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "T-method-L56"
   ]
  },
  {
   "id": "F-1257",
   "kind": "fact",
   "claim": "Sizing a future v0.3: a one-sided \">0.50\" at n 110 needs a point ≈ 0.53 and passes only ≈ 30–50 % of the time; 240–280 items for 80 % power",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "plan",
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 14:24",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "F-chron-late-16"
   ]
  },
  {
   "id": "F-1258",
   "kind": "event",
   "claim": "Team's closing-sprint deadline set to 2026-10-20; closing sprint",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "T-results-L63"
   ]
  },
  {
   "id": "F-1259",
   "kind": "event",
   "claim": "LONGSESSION: no quality drop across 5 turns (and to turn 15)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": "boar-research:arms/longsession/LONGSESSION.md:5,68",
   "caveat": "citation fix: l.3 is the health caveat line",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "T-results-L65"
   ]
  },
  {
   "id": "F-1260",
   "kind": "event",
   "claim": "Team's closing-sprint deadline set to 2026-10-20 (the bounty page says final decision by 2026-10-31); closing sprint",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 21:59",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "T-method-L60"
   ]
  },
  {
   "id": "F-1261",
   "kind": "event",
   "claim": "LONGSESSION: no quality drop in 5 turns, then out to turn 15; cost is latency",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 22:21 / 22:43",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "E-chron-late-L72"
   ]
  },
  {
   "id": "F-1262",
   "kind": "decision",
   "claim": "Porting order for the native app: FACTS + first-aid cards first, then calculator, then grounding rules (why: breakdown of the native-vs-cut7 gap by mechanism)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": "outcome: port spec + mechanism shares sent; relative numbers only public",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "D-system-17"
   ]
  },
  {
   "id": "F-1263",
   "kind": "decision",
   "claim": "Native-app absolute scores stay internal in public text; relative findings only (why: the build was not final; maintainer relationship)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15/17",
   "source_public": null,
   "caveat": "outcome: shared privately with maintainer 10-17; public text holds the number",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "D-system-21"
   ]
  },
  {
   "id": "F-1264",
   "kind": "fact",
   "claim": "Mechanical release gate",
   "value": "9 checks; 1 FAIL (maintainer handle in 2 dataset URLs), fixed by placeholders; one-commit repos with neutral author",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "F-process-72"
   ]
  },
  {
   "id": "F-1268",
   "kind": "event",
   "claim": "Bonsai ternary 9B/4B/2B closed; lite 1.7B chosen as lite candidate",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "T-results-L68"
   ]
  },
  {
   "id": "F-1269",
   "kind": "event",
   "claim": "Submission vehicle = maintainer's native Kotlin app; PR #85 (cut 7 engine as reference) open",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": "no line was given; the 10-15 decision is logged at QUEUE l.1154 (11:00)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "T-results-L69"
   ]
  },
  {
   "id": "F-1270",
   "kind": "event",
   "claim": "Port spec of the health path written for the native app",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "T-system-L61"
   ]
  },
  {
   "id": "F-1271",
   "kind": "event",
   "claim": "PR #85 opened to the upstream app (engine + wiring)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 10:10",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "E-chron-late-L77"
   ]
  },
  {
   "id": "F-1272",
   "kind": "fact",
   "claim": "PR #85 contents at opening",
   "value": "13 cut 7 commits rebased + 1 test + 2 wiring commits; vitest 1,378/1,378; replay 60/60; generation 59/60",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.1 public",
   "set_status": null,
   "judge": "none",
   "date": "2026-10-15",
   "date_text": "2026-10-15 10:10",
   "source_public": "boar-research:arms/pr1/PR1-REBASE.md",
   "caveat": "later + CodeRabbit fixes, NUMGUARD r3, HEALTHGAP r2 (tip 8ac9eb7a)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "F-chron-late-61"
   ]
  },
  {
   "id": "F-1273",
   "kind": "event",
   "claim": "Maintainer: official submission = his native Kotlin app",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 11:00",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "E-chron-late-L79"
   ],
   "superseded_by": [
    "F-1703",
    "F-1686"
   ],
   "superseded_note": "The submission changed on 2026-10-24: BoarLite, a new closed app by the author, replaces the app maintainer's native app as the bounty submission."
  },
  {
   "id": "F-1274",
   "kind": "event",
   "claim": "Official submission vehicle = maintainer's native Kotlin app",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 11:00",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "T-system-L60"
   ],
   "superseded_by": [
    "F-1703",
    "F-1686"
   ],
   "superseded_note": "The submission changed on 2026-10-24: BoarLite, a new closed app by the author, replaces the app maintainer's native app as the bounty submission."
  },
  {
   "id": "F-1275",
   "kind": "decision",
   "claim": "Reject the lexicon numeric guard; ship calculator-only NUMGUARD r3 — why: confirmation review: a lexicon guard cannot be made safe (numeric health questions leak) — outcome: r3 pushed to PR #85",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 21:34–21:56",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "D-chron-late-32"
   ]
  },
  {
   "id": "F-1276",
   "kind": "event",
   "claim": "Ternary Bonsai models closed; Qwen3-1.7B becomes the lite candidate",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 22:14–23:14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "E-chron-late-L85"
   ]
  },
  {
   "id": "F-1277",
   "kind": "event",
   "claim": "Author: lite tier is an open, post-10-20 exploration line; starts with an unsafe-answer warning",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 22:2x / 23:0x",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "E-chron-late-L86"
   ],
   "superseded_by": [
    "F-1386",
    "F-1349"
   ],
   "superseded_note": "Superseded on 2026-10-20: Qwen3-1.7B became the main 8 GB engine (not only a lite tier for 4-6 GB phones); the 4B became the optional tier."
  },
  {
   "id": "F-1278",
   "kind": "decision",
   "claim": "NUMGUARD lexicon guard rejected as unsafe; calculator-only r3 kept (why: a lexicon that lets the model answer a detector-fired health turn cannot be made safe)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "outcome: r3 shipped to PR #85",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "D-results-16"
   ]
  },
  {
   "id": "F-1279",
   "kind": "fact",
   "claim": "Retracted summary: \"the small-model gap is reach, not reading\"",
   "value": "at identical contexts 1.7B − 4B = −0.070 (−0.094…−0.044)",
   "ci": "n 97, 1 seed",
   "n": "97,",
   "seeds": "1",
   "set": "180 self-tests (dev)",
   "set_status": "dev",
   "judge": "Haiku",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "retracted in QUEUE l.1267",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "F-process-59"
   ]
  },
  {
   "id": "F-1280",
   "kind": "fact",
   "claim": "Lite readers with the fact chunk in context",
   "value": "1.7B 0.511 vs 4B 0.543 (−0.032, −0.070…+0.005); LFM2-1.2B −0.060 and Gemma-3-1B −0.065 vs 1.7B",
   "ci": "n 20",
   "n": "20",
   "seeds": null,
   "set": "180 slice dev",
   "set_status": "dev",
   "judge": "Haiku",
   "date": "2026-10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "F-results-93"
   ]
  },
  {
   "id": "F-1281",
   "kind": "fact",
   "claim": "Prefill/decode speed 1.7B vs 4B on M4 CPU",
   "value": "prefill 2.45–2.56×; decode 2.21×",
   "ci": "10 prompts",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": "end-to-end at matched quality ≈ 1.2×",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "F-results-94"
   ]
  },
  {
   "id": "F-1282",
   "kind": "event",
   "claim": "v0.3 write kit handed to the author (items still unwritten)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "T-method-L62"
   ],
   "superseded_by": [
    "F-1374",
    "F-2070"
   ],
   "superseded_note": "No v0.3 was written (decision D5, 2026-10-18); the v0.4 ruler set (half Claude, half OpenAI gpt-5.6-sol, frozen 2026-10-20) took its place and is not written outside the Claude family."
  },
  {
   "id": "F-1283",
   "kind": "event",
   "claim": "NUMGUARD r3 + HEALTHGAP r2 added to the upstream PR; WARN banner spec v2",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "T-system-L63"
   ]
  },
  {
   "id": "F-1284",
   "kind": "decision",
   "claim": "TK2-FIX is a correctness fix, not a score lever; out of PR #85 before 10-20 — why: +0.010 pooled, needs a builder change too — outcome: queued as a post-10-20 patch",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 03:39",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "D-chron-late-34"
   ]
  },
  {
   "id": "F-1285",
   "kind": "event",
   "claim": "NUMGUARD r3 + HEALTHGAP r2 on PR #85",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-16",
   "date_text": "2026-10-16 09:38",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "E-chron-late-L90"
   ]
  },
  {
   "id": "F-1286",
   "kind": "fact",
   "claim": "Lite safety check",
   "value": "deterministic layers answered 61/64 probes, so the 1.7B answered 3; unsafe 1 vs 1",
   "ci": "n 12 effective",
   "n": "12",
   "seeds": null,
   "set": "agent probes",
   "set_status": null,
   "judge": "jev + agent",
   "date": "2026-10-16",
   "date_text": "2026-10-16 10:08",
   "source_public": null,
   "caveat": "almost no power: \"routing 1.7B = 4B 124/124\" is the safe wording",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "F-chron-late-70"
   ]
  },
  {
   "id": "F-1289",
   "kind": "fact",
   "claim": "The native app does not run the research engine's health path or deterministic answers.",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "source_public": null,
   "caveat": "from maintainer's own portal report + guard texts missing",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "F-system-84"
   ]
  },
  {
   "id": "F-1290",
   "kind": "event",
   "claim": "Submitted app (maintainer's native build) scored for the first time; maintainer-supplied traces",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "date_text": "2026-10-17 04:05, 12:22",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "E-process-L74"
   ]
  },
  {
   "id": "F-1291",
   "kind": "event",
   "claim": "Submitted native app scored internally on 57 public items (numbers held internal)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "date_text": "2026-10-17 04:05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "T-results-L75"
   ]
  },
  {
   "id": "F-1292",
   "kind": "decision",
   "claim": "Claim v4.2.4: state that the submitted app ports retrieval and prompts, not the health path or deterministic answers — why: maintainer's own portal report + MAINT-V01 — outcome: final claim wording",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "date_text": "2026-10-17 20:59",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "D-chron-late-40"
   ]
  },
  {
   "id": "F-1293",
   "kind": "event",
   "claim": "Claim wording corrected: native app ports retrieval and prompts, not the health path or deterministic answers",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "date_text": "2026-10-17 20:59",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "T-system-L68"
   ]
  },
  {
   "id": "F-1295",
   "kind": "event",
   "claim": "HONESTPROMPT launched (maintainer's honesty-prompt edits vs invented facts)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-17",
   "date_text": "2026-10-17 23:07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "E-chron-late-L102"
   ]
  },
  {
   "id": "F-1296",
   "kind": "event",
   "claim": "Bounty deadline (moved)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "T-system-L69"
   ]
  },
  {
   "id": "F-1297",
   "kind": "fact",
   "claim": "v0.3 similarity gate: 0 pairs with word-Jaccard > 0.30 or embedding cosine ≥ 0.90 against public, safety and every held-out pool.",
   "value": "thresholds",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.3",
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-bench:CONTRIBUTING.md:30",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "F-method-87"
   ]
  },
  {
   "id": "F-1298",
   "kind": "null-result",
   "claim": "LITE4 source-bound rule: force the 1.7B to answer only from sources → −0.001 (−0.064…+0.063) (flat)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "N-chron-late-L291"
   ]
  },
  {
   "id": "F-1299",
   "kind": "null-result",
   "claim": "LITE8 fact selection: target the 1.7B's not-read bucket → ceiling ≈ +0.014 < MDE (closed by ceiling rule)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "N-chron-late-L293"
   ]
  },
  {
   "id": "F-1300",
   "kind": "null-result",
   "claim": "QUANTLITE: Q8_0 vs Q4_K_M for the 1.7B → −0.003 (−0.025…+0.019); decode 0.68× (flat)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "N-chron-late-L295"
   ]
  },
  {
   "id": "F-1301",
   "kind": "null-result",
   "claim": "LITEREADERS: other ≤ 1.2 GB readers on identical contexts → LFM2-1.2B −0.060, Gemma-3-1B −0.065 vs 1.7B (1.7B best of the small readers)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "N-chron-late-L296"
   ]
  },
  {
   "id": "F-1302",
   "kind": "null-result",
   "claim": "R83 Ternary 2B vs stock 1.7B: 120 Haiku",
   "value": "−0.057 (−0.104…−0.014) kit",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "R83 LOSS on n 20 slices; CRITIQUE-2026-10-16.md:59 lists it among closures that should read 'not detected at MDE 0.08'",
   "verification": "OVERCLAIM",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "N-results-L304"
   ]
  },
  {
   "id": "F-1303",
   "kind": "null-result",
   "claim": "R97 Thinking ON for lite 1.7B: slice 20",
   "value": "512: −0.041 (−0.084…+0.003); 256: −0.001",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "R97 n 20, CI upper +0.003 at 512 / +0.038 at 256: 'not detected at MDE 0.08' per CRITIQUE-2026-10-16.md:59",
   "verification": "OVERCLAIM",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "N-results-L318"
   ]
  },
  {
   "id": "F-1304",
   "kind": "null-result",
   "claim": "Lite readers: LFM2-1.2B, Gemma-3-1B",
   "value": "−0.060 / −0.065 vs 1.7B",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "why closed: 1.7B best",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "N-system-L222"
   ]
  },
  {
   "id": "F-1305",
   "kind": "decision",
   "claim": "Build a thin own harness (protocol + trace schema + adapters + existing scorers) instead of adopting an eval framework (why: scorers already existed and were tested; the engine is TypeScript with 3 runtimes)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-01",
   "source_public": null,
   "caveat": "outcome: boar-eval became the harness; later the public boar-bench",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "reproduce",
   "aliases": [
    "D-method-01"
   ]
  },
  {
   "id": "F-1306",
   "kind": "decision",
   "claim": "Primary metric = raw ratio, bar > 0.50 (the author's decision on 2026-10-02, not the app maintainer's)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-02",
   "source_public": "boar-bench:bench/v0.3/plan/analysis-plan.md:17",
   "caveat": "the scrubbed public plan says 'the maintainer's decision', but the private original names the author; in the publication 'the app maintainer' means the upstream maintainer, so write 'the author's decision'",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "reproduce",
   "aliases": [
    "D-method-07"
   ]
  },
  {
   "id": "F-1307",
   "kind": "event",
   "claim": "Pack catalogue + authoring workflow (licence policy PD/CC0/CC BY/CC BY-SA only)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-12",
   "date_text": "2026-10-12 22:46",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "reproduce",
   "aliases": [
    "E-chron-late-L45"
   ]
  },
  {
   "id": "F-1308",
   "kind": "fact",
   "claim": "Pack format 2.",
   "value": "one SQLite file; article text in zstd blocks; 1,000-char chunks with FTS; one lead vector (first 2,500 chars) per article; per-row licence + attribution",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-13",
   "source_public": "boar-research:reports/REPORT-PACKS-2026-10-13.md:27",
   "caveat": "format v3 adds chunk vectors (≈ +60 % size)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "reproduce",
   "aliases": [
    "F-system-63"
   ]
  },
  {
   "id": "F-1309",
   "kind": "fact",
   "claim": "Release hazard scan before scrubbing",
   "value": "research: 69 local-path lines, 109 handle/name lines, 30 blind-path mentions, 0 held-out text lines",
   "ci": "411 files",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "source_public": null,
   "caveat": "heuristic scan",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "reproduce",
   "aliases": [
    "F-process-73"
   ]
  },
  {
   "id": "F-1310",
   "kind": "decision",
   "claim": "Publish research AND benchmark for other teams; publish v0.1 reference-answer text; v0.2/V-para refs stay hashes only — why: other teams cannot run the judge comparably without refs — outcome: boar-bench carries refs with provenance on the dataset card",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 22:17–22:20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "reproduce",
   "aliases": [
    "D-chron-late-27"
   ]
  },
  {
   "id": "F-1311",
   "kind": "fact",
   "claim": "Release scrub hazards found and removed before publishing: absolute home-directory paths, the author's handle, blind/veil path strings and held-out ids; no held-out item text was found.",
   "value": "[local path] paths 23 + 69; handle 29 + 109; blind/veil path strings 22 + 30; held-out ids 8 + 19; no held-out item text found",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-14",
   "date_text": "2026-10-14 22:29",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "reproduce",
   "aliases": [
    "F-chron-late-60"
   ]
  },
  {
   "id": "F-1312",
   "kind": "decision",
   "claim": "Release as two fresh repos with one squashed commit, neutral author, curated whitelist + scrub + mechanical gate + independent review; never the private history — why: Private history holds held-out plaintext and personal data — outcome: Gate FAIL on 2 handle URLs fixed; 41/44 review findings applied; live 10:07",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "reproduce",
   "aliases": [
    "D-process-24"
   ]
  },
  {
   "id": "F-1313",
   "kind": "fact",
   "claim": "The emergency detector is an mmBERT-small encoder (q8_0) with a small classification head.",
   "value": "106,210,784 B GGUF + 1,696,707 B head; threshold 0.2, floor 0.01",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "source_public": "boar-research:BOM.md:17",
   "caveat": "not public; base-model licence UNKNOWN",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "reproduce",
   "aliases": [
    "F-system-04"
   ]
  },
  {
   "id": "F-1314",
   "kind": "event",
   "claim": "Claim v4.0 reviewed; public repos staged and gated (41/44 review findings applied)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 01:03–01:18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "reproduce",
   "aliases": [
    "E-chron-late-L75"
   ]
  },
  {
   "id": "F-1315",
   "kind": "event",
   "claim": "Public repos staged: scrub, mechanical gate, independent review (41/44 findings applied)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 01:18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "reproduce",
   "aliases": [
    "E-process-L62"
   ]
  },
  {
   "id": "F-1316",
   "kind": "fact",
   "claim": "Public repos at release",
   "value": "boar-bench 139 files / 2.26 MB; boar-research 411 files / 5.53 MB",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 01:18 / 10:07",
   "source_public": null,
   "caveat": "one commit each, author \"BOAR R&D\"",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "reproduce",
   "aliases": [
    "F-chron-late-59"
   ]
  },
  {
   "id": "F-1317",
   "kind": "event",
   "claim": "Public bench and research repos staged and verified (scrubbed, reviewed)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 01:18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "reproduce",
   "aliases": [
    "T-method-L61"
   ]
  },
  {
   "id": "F-1318",
   "kind": "event",
   "claim": "Public repos live: boar-bench (139 files) and boar-research (411 files)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-15",
   "date_text": "2026-10-15 10:07",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "reproduce",
   "aliases": [
    "E-chron-late-L76"
   ]
  },
  {
   "id": "F-1319",
   "kind": "fact",
   "claim": "The Python stats port reproduces the JS bootstrap bit-for-bit (tested against node).",
   "value": "bit-identical",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": "boar-bench:scorers/stats.py:1-3",
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "reproduce",
   "aliases": [
    "F-method-19"
   ]
  },
  {
   "id": "F-1320",
   "kind": "event",
   "claim": "Literature pass on prompting 1-8B models for honesty without over-refusal (PROMPT-LIT), Haiku sweep + Opus verification",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-01"
   ]
  },
  {
   "id": "F-1321",
   "kind": "event",
   "claim": "Publication reframed as living preprint v0.1 \"as of 2026-10-18\" (internal build state)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-02"
   ]
  },
  {
   "id": "F-1322",
   "kind": "event",
   "claim": "HONESTPROMPT set-A audit: honest Detailed prompt cuts wrong claims 0.231 -> 0.062",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-03"
   ]
  },
  {
   "id": "F-1323",
   "kind": "event",
   "claim": "HONESTPROMPT verdict: GO for Detailed honest (D1), NO for Succinct-honest (S1)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "T-u-1820-04"
   ]
  },
  {
   "id": "F-1324",
   "kind": "event",
   "claim": "HONESTPROMPT closed; capability sentence does not make Detailed write code (route code asks instead)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-05"
   ]
  },
  {
   "id": "F-1325",
   "kind": "event",
   "claim": "CHATFAIL S1: routing fix for code/format asks PARTIAL on 4B (+0.122 detector, ~+0.10 audited), NO on 1.5B",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-06"
   ]
  },
  {
   "id": "F-1326",
   "kind": "event",
   "claim": "D4 prompt (D1 + \"always do what the user asks\" capability rule) = D1 on v0.1 (+0.003); D4 becomes recommended Detailed prompt",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-07"
   ]
  },
  {
   "id": "F-1328",
   "kind": "event",
   "claim": "NEXT-STEPS decision document: 7 open decisions, three closure paths, and a conflict between the submission date and a later app release date",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "T-u-1820-09"
   ]
  },
  {
   "id": "F-1329",
   "kind": "event",
   "claim": "Author delegates D1-D7; adopted as recommended (no v0.3; submit with claim v4.2.4; site v1.0 only at closure)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-1820-10"
   ]
  },
  {
   "id": "F-1330",
   "kind": "event",
   "claim": "Three engine bug fixes staged from the maintainer's plugin suite: quadratic digit regex, CO card \"in the car\", facts table currency/QID labels (r7)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "date_text": "2026-10-18 16:59-18:48",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "T-u-1820-11"
   ]
  },
  {
   "id": "F-1331",
   "kind": "event",
   "claim": "FACTSAUDIT: facts r7 vs CIA Factbook, 5 table-wrong rows; r8 fixes + emergency numbers for 195 countries",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-1820-12"
   ]
  },
  {
   "id": "F-1332",
   "kind": "event",
   "claim": "PROMPT-LIT-2 verified (19 claims): small models misuse offered tools; LFM2.5 card says not for knowledge QA without retrieval",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "date_text": "2026-10-19 13:15-13:20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-1820-13"
   ]
  },
  {
   "id": "F-1333",
   "kind": "event",
   "claim": "Author redirects to \"tests with sources\" (user has the matching pack): PACKREAD launched",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-1820-14"
   ]
  },
  {
   "id": "F-1334",
   "kind": "event",
   "claim": "SAFEPACK: emergency pack installed, engine is safe but unhelpful (SAFE-FAIL 26/30)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-1820-15"
   ]
  },
  {
   "id": "F-1335",
   "kind": "event",
   "claim": "SAFEROUTE fix raises gold-page reach 1 -> 21/30 but adds unsafe verbatim extracts; frozen rule FAIL kept on record, author releases the patch anyway (later held)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "date_text": "2026-10-19 14:12-14:27",
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-16"
   ]
  },
  {
   "id": "F-1336",
   "kind": "event",
   "claim": "SMALLCLS: 1.7B hazard classifier; review later shows it names the right hazard on only 14/45",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-17"
   ]
  },
  {
   "id": "F-1337",
   "kind": "event",
   "claim": "Independent direction review: stop opening arms, consolidate, hold saferoute",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-18"
   ]
  },
  {
   "id": "F-1338",
   "kind": "event",
   "claim": "PACKREAD complete: 4B 0.650 > LFM2.5 0.554 > 1.7B 0.529 reading packs (D4 prompt)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-1820-19"
   ]
  },
  {
   "id": "F-1339",
   "kind": "event",
   "claim": "Engine PR updated with 8 commits (saferoute held), 1,586/1,586 tests",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-20"
   ]
  },
  {
   "id": "F-1340",
   "kind": "event",
   "claim": "RELGATE: 4B picks the right source 72 %; per-chunk relevance and sufficiency gates FAIL",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-1820-21"
   ]
  },
  {
   "id": "F-1341",
   "kind": "event",
   "claim": "V-para diagnosis: loss split wrong 0.309 / absent 0.239 / thin 0.038; little addressable on 8 GB",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-23"
   ]
  },
  {
   "id": "F-1342",
   "kind": "event",
   "claim": "PACKGAP plan: no new open corpus likely to move V-para (<= +0.01)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-1820-24"
   ]
  },
  {
   "id": "F-1343",
   "kind": "event",
   "claim": "BENCH-RULER synthesis: five misreadings of the bounty in the current bench; new public ruler design",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-1820-25"
   ]
  },
  {
   "id": "F-1344",
   "kind": "event",
   "claim": "NIST-crypto pack and visa pack built; both inert on self-tests (+0.008, +0.007)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "date_text": "2026-10-19 23:50 / 10-20 00:14",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-1820-26"
   ]
  },
  {
   "id": "F-1345",
   "kind": "event",
   "claim": "CHAINWHEN NO: 4B source-pick before the answer +0.019",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-27"
   ]
  },
  {
   "id": "F-1346",
   "kind": "event",
   "claim": "V-exact acceptance harness baseline 0.475 (Haiku, 2 seeds), venues 6/6",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-28"
   ]
  },
  {
   "id": "F-1347",
   "kind": "event",
   "claim": "v0.4 ruler candidate: 70 items (Claude 36, OpenAI gpt-5.6-sol 34), 5 disputed points for the author",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-1820-29"
   ]
  },
  {
   "id": "F-1348",
   "kind": "event",
   "claim": "LITE17 NO: thinking, quote-then-answer, chunk filtering leave the 1.7B at ~0.50 vs 4B ~0.61",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-1820-30"
   ]
  },
  {
   "id": "F-1349",
   "kind": "event",
   "claim": "Pivot: Qwen3-1.7B becomes the main 8 GB engine after Pixel 8a speed runs; LITE-MAIN program opened",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-31"
   ]
  },
  {
   "id": "F-1350",
   "kind": "event",
   "claim": "LITEDECLINE PARTIAL: engine's compact-tier declines; releasing all five raises U0 +0.170 but doubles confident errors (17 -> 34)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-32"
   ]
  },
  {
   "id": "F-1351",
   "kind": "event",
   "claim": "LITELAT on Pixel 8a: retrieval is 57 % of time to first token (~435 serial SQL calls)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-1820-33"
   ]
  },
  {
   "id": "F-1352",
   "kind": "event",
   "claim": "CMP30 / CMP30OFF on the Pixel 8a: 1.7B vs Qwen2.5-1.5B; with declines released 1.7B leads (correctness +0.46)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "date_text": "2026-10-20 16:14 / 16:55",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-34"
   ]
  },
  {
   "id": "F-1353",
   "kind": "event",
   "claim": "v0.4 ruler item set FROZEN (75 items, sha f9ad2f2f...) before any system answered",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-1820-35"
   ]
  },
  {
   "id": "F-1354",
   "kind": "event",
   "claim": "BATCHRET: batched pack SQL (937 -> 48 calls, byte-identical retrieval 390/390); on the phone TTFT 9.12 -> 5.55 s",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "date_text": "2026-10-20 16:48 / 17:54",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-1820-36"
   ]
  },
  {
   "id": "F-1355",
   "kind": "event",
   "claim": "LITEPACKS NO CHANGE on MAIN; packs help only pack-matched self-tests",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "Verification note: Time/source: MAIN NO CHANGE verdict was computed 18:23 ; is the 21:12 final report with the secondary sets.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-1820-37"
   ]
  },
  {
   "id": "F-1356",
   "kind": "event",
   "claim": "Author \"grill\": 1.7B stays default, 4B optional tier for 10-12 GB phones, two-stage screens, health weight 0.5 in the ruler, safety-cards-first work archived",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-1820-38"
   ]
  },
  {
   "id": "F-1357",
   "kind": "event",
   "claim": "Five 1.7B levers NO-GO at screen: honest prompt, short context, quote-first, greedy, vendor sampling; token confidence (AUC 0.53) and first-token entropy (0.49) do not flag errors",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "date_text": "2026-10-20 19:39-19:56",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-39"
   ]
  },
  {
   "id": "F-1358",
   "kind": "event",
   "claim": "CE taxonomy: 20 of 27 1.7B confident errors had no relevant source",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-40"
   ]
  },
  {
   "id": "F-1359",
   "kind": "event",
   "claim": "Author: ruler split into with-reference / no-reference parts; product refuses by default with \"show anyway\"; persona-pack arms",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-1820-41"
   ]
  },
  {
   "id": "F-1360",
   "kind": "event",
   "claim": "Ruler spec v1 r2 written + independent adversarial review (25 findings); verdict arm = union of persona packs",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-1820-42"
   ]
  },
  {
   "id": "F-1361",
   "kind": "event",
   "claim": "Model-usage audit: 33 findings, 6 P0 (stale CE metric, refusals counted as CE, unscreened candidate model)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-43"
   ]
  },
  {
   "id": "F-1362",
   "kind": "event",
   "claim": "REREAD: in refuse-by-default mode each old screen held only 4 visible confident errors; the 6-fix bar was unreachable; quote-first is a regression",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-44"
   ]
  },
  {
   "id": "F-1363",
   "kind": "event",
   "claim": "Groundedness add-on gate (GATE17-SUPPORT, AUC 0.965) and REFUSE-UNSUPPORTED PASS: visible confident errors 7 -> 3 of 177 under the Claude Opus judge (10 -> 4 under Claude Haiku); +9.3 s p50 projected for the Pixel 8a (5.1 s measured on a desktop)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "date_text": "2026-10-20 21:17 / 22:09",
   "source_public": null,
   "caveat": "Verification note: '+9 s on phone' is a projection (Pixel 8a prefill 9.3 s p50, measured 5.1 s on a desktop), not a phone measurement. Under Opus the headline is 7 -> 3 of 177 (CE-OPUS-RU, 10-21 03:35), Haiku 10 -> 4.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-45"
   ]
  },
  {
   "id": "F-1364",
   "kind": "event",
   "claim": "Persona map v1 drafted + independently approved (not frozen; waits author veto)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-1820-46"
   ]
  },
  {
   "id": "F-1365",
   "kind": "event",
   "claim": "Judge v2 (binary key-point checklist) NO-GO by frozen test-retest rule",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-47"
   ]
  },
  {
   "id": "F-1366",
   "kind": "event",
   "claim": "GATE v2: screens recalibrated against measured null runs (null alone fixes 23-31 %)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-48"
   ]
  },
  {
   "id": "F-1367",
   "kind": "event",
   "claim": "Gemma 4 E2B: GO on all four screens, stage 2 FAIL by 0.002",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-49"
   ]
  },
  {
   "id": "F-1368",
   "kind": "event",
   "claim": "Overnight model screens: Q8 quant, Qwen3.5-2B, OCC-RAG, FOTAPOL LoRA, Jan-v1-edge, SmolLM3, granite-1B all NO; Ministral-3-3B + citation prompt stage-2 PASS at ~3x phone time",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "date_text": "2026-10-21 00:00-02:36",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-50"
   ]
  },
  {
   "id": "F-1369",
   "kind": "event",
   "claim": "Gemma 4 E2B on the Pixel 8a: 73/73, answer p50 11.8 s; judged = desktop, 0.083 below the 4B on the same phone",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "date_text": "2026-10-21 02:23-02:30",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-1820-51"
   ]
  },
  {
   "id": "F-1370",
   "kind": "decision",
   "claim": "D1: no LFM2.5 desktop replay",
   "value": "closed",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "who decided: delegated coordinator (author delegated); why: audit already 7/58 vs 3/73 wrong; product already chose Qwen; alternatives: run replay ~1.5 h; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-1820-01"
   ]
  },
  {
   "id": "F-1371",
   "kind": "decision",
   "claim": "D2: submit on the 10-20 date with claim v4.2.4 (app \"scored internally, final build published when it ships\")",
   "value": "post prepared ; posting not recorded in slice",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "who decided: delegated coordinator; why: meets date; numbers honest; alternatives: ask extension to a later date; submit research build; reversible: partly | Verification note: The parenthetical wording ('scored internally, final build published when it ships') is the claim text's, sourced at not .",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-1820-02"
   ]
  },
  {
   "id": "F-1372",
   "kind": "decision",
   "claim": "D3: publish claim text now rather than hold",
   "value": "paste-ready 1,102 words",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "who decided: delegated coordinator; why: final and reviewed; alternatives: hold to 10-20; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-1820-03"
   ]
  },
  {
   "id": "F-1373",
   "kind": "decision",
   "claim": "D4: one deep-tier session on the 12 GB phone only if in hand",
   "value": "phone location asked; not run in slice",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "who decided: delegated coordinator; why: closes the phone front either way; alternatives: record \"desktop only\"; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-1820-04"
   ]
  },
  {
   "id": "F-1374",
   "kind": "decision",
   "claim": "D5: no bench v0.3 (non-Claude held-out set) before closure",
   "value": "record declares no unexposed set; later partly superseded by v0.4 ruler (see §6)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "who decided: delegated coordinator; why: ~14-24 h of author item writing; would land after the deadline; bar needs ~0.53 at n 110 vs best ~0.43-0.51; serves honesty, not winning; alternatives: 60-item wave 1; 6-item pilot; reversible: yes | VERIFIER NOTE: Base OF-3 says 'the author decided'; precise form: the author delegated D1-D7 and the coordinator's recommendation (skip v0.3) was adopted ( l.4).",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-1820-05"
   ]
  },
  {
   "id": "F-1375",
   "kind": "decision",
   "claim": "D6: archive the native-app v0.1 note (do not send)",
   "value": "archived",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "who decided: delegated coordinator; why: content already reached the maintainer in parts; alternatives: send via liaison; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-1820-06"
   ]
  },
  {
   "id": "F-1376",
   "kind": "decision",
   "claim": "D7: upload the research site only as v1.0 at closure; v0.x are internal build states",
   "value": "folded into publication PLAN",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "who decided: delegated coordinator; why: one publication, no intermediate states; alternatives: upload v0.1 now; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-1820-07"
   ]
  },
  {
   "id": "F-1377",
   "kind": "decision",
   "claim": "Recommend the honest Detailed prompt D1, then D4 (D1 + capability rule) as the Detailed prompt; no Succinct-honest",
   "value": "sent to maintainer via liaison",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "date_text": "2026-10-18 05:02-10:30",
   "source_public": null,
   "caveat": "who decided: coordinator under frozen rule; why: wrong claims down, ratio within -0.03 rule, refusals 1-2/49; S1 refuses 7/49; alternatives: D2, S1, S3; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-1820-08"
   ]
  },
  {
   "id": "F-1378",
   "kind": "decision",
   "claim": "Never ship a capability sentence without health forced to guard texts",
   "value": "rule standing",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "who decided: coordinator; why: L1 line added wrong first-aid steps on first-aid chats; alternatives: ship L1; reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-1820-09"
   ]
  },
  {
   "id": "F-1379",
   "kind": "decision",
   "claim": "Focus finishing work on \"tests with sources\" (user has the matching pack)",
   "value": "PACKREAD run; PROMPT3M paused, later closed",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "who decided: author; why: bounty finish; alternatives: continue no-source prompt tests (PROMPT3M); reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-1820-10"
   ]
  },
  {
   "id": "F-1380",
   "kind": "decision",
   "claim": "Release saferoute patch despite SAFEPACK frozen-rule FAIL (app ships with a warning); FAIL kept on record",
   "value": "reversed same day: held out of the pushed stack after the direction review",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "who decided: author; why: rule judged too strict for stage; alternatives: hold patch; reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-1820-11"
   ]
  },
  {
   "id": "F-1381",
   "kind": "decision",
   "claim": "Stop opening new arms; consolidate; hold saferoute; correct over-reads (SMALLCLS, SAFEPACK)",
   "value": "adopted for saferoute; new arms nonetheless resumed 10-20 under new program",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "who decided: independent reviewer, adopted by coordinator; why: scope drift two days before closure; small-n in-sample results drove release calls; alternatives: open SMALLCLS-2 / agentic search loop; reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-1820-12"
   ]
  },
  {
   "id": "F-1382",
   "kind": "decision",
   "claim": "Resume three safety-held items: per-answer health caution ON, drinking-water card, chat routing L2+L3 without L1",
   "value": "in the pushed PR stack",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "who decided: author; why: app ships with a warning; harm came from L1 line only; alternatives: keep held; reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-1820-13"
   ]
  },
  {
   "id": "F-1383",
   "kind": "decision",
   "claim": "Build a public ruler: true-zero utility U + task success, uncapped two-family web references, non-Claude items, submitted app on phone, gates, speed, install",
   "value": "led to v0.4 items + spec v1",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "who decided: coordinator from 8-agent workflow (both critics converged); why: five misreadings of the bounty in the old bench; alternatives: keep old bench as headline; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-1820-14"
   ]
  },
  {
   "id": "F-1384",
   "kind": "decision",
   "claim": "Item families from Vitalik Buterin's own public test questions; travel ~1/3 not >= 50 %; 8 multi-turn; over-refusal probes; venue \"best\" checked on open web",
   "value": "applied to v0.4",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "date_text": "2026-10-20 (r2)",
   "source_public": null,
   "caveat": "who decided: coordinator, author direction; why: match the inspiring post's queries; alternatives: r1 quotas (72 items, >= 50 % travel); reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-1820-15"
   ]
  },
  {
   "id": "F-1385",
   "kind": "decision",
   "claim": "Resolve 5 disputed points; drop one item contradicted by both families; tag 4 V-exact anchors out of the headline; freeze v0.4",
   "value": "frozen, sha f9ad2f2f...",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "who decided: author; why: both-family fact check; alternatives: keep disputed items; reversible: no (frozen)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-1820-16"
   ]
  },
  {
   "id": "F-1386",
   "kind": "decision",
   "claim": "Qwen3-1.7B = main 8 GB engine; 4B = optional \"more complete\" tier",
   "value": "adopted; confirmed at 19:07 grill",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "date_text": "2026-10-20 ~11:50",
   "source_public": null,
   "caveat": "who decided: author; why: Pixel 8a: 1.7B 13-16 tok/s, answers 13-16 s; 4B 3.6-6 tok/s, 29-46 s, one LMK kill; alternatives: keep 4B default; Qwen2.5-1.5B (faster but weaker); reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-1820-17"
   ]
  },
  {
   "id": "F-1387",
   "kind": "decision",
   "claim": "Propose to maintainer: 8 GB = Qwen3-1.7B + fixed warning line instead of compact declines",
   "value": "superseded at 20:24 by refuse-by-default",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "date_text": "2026-10-20 ~17:35",
   "source_public": null,
   "caveat": "who decided: author (send OK); why: CMP30OFF: 1.7B with declines released leads 1.5B; alternatives: keep declines; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-1820-18"
   ]
  },
  {
   "id": "F-1388",
   "kind": "decision",
   "claim": "Grill Q1-Q13: three fronts; every arm a two-stage screen (problems + sentinels, then full MAIN 97); safety-cards-first archived; engine changes sent only after a measured gain; Vitalik Buterin-blog pack built user-selected, unmeasured",
   "value": "adopted",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "who decided: author; why: LITEPACKS NO CHANGE; alternatives: phone A/B of 4B; C before bounty; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-1820-19"
   ]
  },
  {
   "id": "F-1389",
   "kind": "decision",
   "claim": "Ruler headline weights the 11 health_safety items at 0.5, adds a safety gate, publishes weight-1 U beside it",
   "value": "in spec v1 §2; reviewer F3 flags timing (decided after items existed)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "who decided: author; why: bounty asks for casual lookup/research, never names health; alternatives: weight 1; reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-1820-20"
   ]
  },
  {
   "id": "F-1390",
   "kind": "decision",
   "claim": "Ruler headline = coverage x U decomposition + U-total; product refuses by default when no source is kept, with a \"show anyway\" control",
   "value": "spec v1 r2",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "who decided: author; why: separate what packs answer from memory answers; alternatives: answer under a warning line; reversible: yes Metric: ruler v0.4 U-total (ratio of sums against the reference mass, 71 headline items, health weight 0.5); never compare with the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-1820-21"
   ]
  },
  {
   "id": "F-1391",
   "kind": "decision",
   "claim": "Verdict arm = A∪ (all persona packs installed once, same for every item); per-family arm A kept as a labelled row",
   "value": "spec v1 r2",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "who decided: author on reviewer finding F1; why: per-family arm = oracle topic routing; alternatives: arm A or arm B as verdict; reversible: yes | Verification note: Reviewer finding F1 is at (TL;DR 1) and :14 (F1 heading); line 7 is the decline-arm finding.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-1820-22"
   ]
  },
  {
   "id": "F-1392",
   "kind": "decision",
   "claim": "Persona map v1: travel -> wikivoyage + visa (+ area tile); tech -> crypto; science -> medlineplus + firstaid; place -> factbook; humanities -> none; NIST pack excluded",
   "value": "not frozen, waits author veto",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "who decided: coordinator draft + independent approver (6 edits); why: chosen from pack-card domain lines only; alternatives: include NIST, simplewiki; reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-1820-23"
   ]
  },
  {
   "id": "F-1393",
   "kind": "decision",
   "claim": "GATE v2: screen bar = max measured null net + 2",
   "value": "used for overnight screens",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "who decided: coordinator; why: null runs fixed 23-31 % of problems; old 6-fix bar sat at the null; alternatives: fixed bar 6; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-1820-24"
   ]
  },
  {
   "id": "F-1394",
   "kind": "fact",
   "claim": "Honest Detailed prompt (D1) cuts wrong-claim rate on the app's chat-template path",
   "value": "0.231 -> 0.062 (paired -0.169)",
   "ci": "CI -0.281..-0.073; n 15 answers/arm, 407 claims",
   "n": "15",
   "seeds": null,
   "set": "set A (agent-written chat turns), dev",
   "set_status": "dev",
   "judge": "Claude Opus sentence audit; wrong claims / claims",
   "date": "2026-10-18",
   "date_text": "10-18",
   "source_public": null,
   "caveat": "answers 42 % shorter; vaguer on 2/15",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-01"
   ]
  },
  {
   "id": "F-1395",
   "kind": "fact",
   "claim": "Pooled A+B wrong-claim rate D0 vs D1",
   "value": "0.216 -> 0.098",
   "ci": "35 answers/arm",
   "n": "35",
   "seeds": null,
   "set": "sets A+B, dev",
   "set_status": "dev",
   "judge": "Claude Opus audit",
   "date": "2026-10-18",
   "date_text": "10-18",
   "source_public": null,
   "caveat": "crypto errors persist (knowledge)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-02"
   ]
  },
  {
   "id": "F-1396",
   "kind": "fact",
   "claim": "D4 prompt keeps the honesty gain at full length",
   "value": "wrong 0.065 at 11.2 claims/answer (D0 length); D4-D0 -0.166",
   "ci": "CI -0.270..-0.092; n 15",
   "n": "15",
   "seeds": null,
   "set": "set A, dev",
   "set_status": "dev",
   "judge": "Claude Opus audit",
   "date": "2026-10-18",
   "date_text": "10-18",
   "source_public": null,
   "caveat": "single seed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-03"
   ]
  },
  {
   "id": "F-1397",
   "kind": "fact",
   "claim": "D4 costs no judged ratio on v0.1",
   "value": "D4 0.427 vs D1 0.425 (+0.003); D0 0.452",
   "ci": "CI -0.024..+0.028; 57 ids, 1 seed",
   "n": "57",
   "seeds": "1",
   "set": "v0.1 public dev",
   "set_status": "dev",
   "judge": "Claude Haiku 5.5-low, raw ratio",
   "date": "2026-10-18",
   "date_text": "10-18",
   "source_public": null,
   "caveat": "win 0/57 for both. Proxy contexts (not the app retrieval), one seed, Claude Haiku judge; raw ratio only (the source has no floor-adjusted value).",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-04"
   ]
  },
  {
   "id": "F-1398",
   "kind": "fact",
   "claim": "Refusals on capability turns",
   "value": "D0 1/49, D1 2/49, D4 1/49, S1 7/49",
   "ci": "49 capability turns",
   "n": null,
   "seeds": null,
   "set": "CHAT-FU-v1, dev",
   "set_status": "dev",
   "judge": "count",
   "date": "2026-10-18",
   "date_text": "10-18",
   "source_public": null,
   "caveat": "The original Detailed prompt D0 wrote code blocks on 8/21 code asks; the honest variants D1/D3/D4 answer in prose (0-2/21).",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-05"
   ]
  },
  {
   "id": "F-1400",
   "kind": "fact",
   "claim": "Chat routing fix L2+L3 on 4B",
   "value": "fulfilled 0.633 -> 0.755 (+0.122) detector; ~+0.10 real after audit",
   "ci": "CI 0.000..+0.256; 49 turns",
   "n": "49",
   "seeds": null,
   "set": "CHAT-FU-v1, dev",
   "set_status": "dev",
   "judge": "detector + Claude Opus audit",
   "date": "2026-10-18",
   "date_text": "10-18",
   "source_public": null,
   "caveat": "1/20 code answers runnable",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-07"
   ]
  },
  {
   "id": "F-1401",
   "kind": "fact",
   "claim": "PACKREAD: reading with packs, D4 prompt",
   "value": "4B 0.650, LFM2.5 0.554, 1.7B 0.529; engine Succinct anchor 0.610",
   "ci": "54 pack self-test items; LF-D4 vs M4-D4 -0.096 (-0.145..-0.052)",
   "n": null,
   "seeds": null,
   "set": "pack self-tests, dev",
   "set_status": "dev",
   "judge": "Claude Haiku, raw ratio",
   "date": "2026-10-19",
   "date_text": "10-19",
   "source_public": null,
   "caveat": "identical contexts: 1.7B gap is reading",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-08"
   ]
  },
  {
   "id": "F-1402",
   "kind": "fact",
   "claim": "SAFEPACK, emergency pack installed, engine as-is",
   "value": "unsafe 0, declined 20, SAFE-FAIL 26/30",
   "ci": "n 30",
   "n": "30",
   "seeds": null,
   "set": "safepack (agent-written Ready.gov/FM items), dev",
   "set_status": "dev",
   "judge": "blind Claude Opus audit",
   "date": "2026-10-19",
   "date_text": "10-19",
   "source_public": null,
   "caveat": "\"safe but not helpful\" AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-09"
   ]
  },
  {
   "id": "F-1403",
   "kind": "fact",
   "claim": "SAFEROUTE patch",
   "value": "gold page in sources 1 -> 21/30; unsafe 0 -> 3 (review: 2 caused + 1 audit split)",
   "ci": "n 30, in-sample",
   "n": "30",
   "seeds": null,
   "set": "safepack, dev",
   "set_status": "dev",
   "judge": "blind Claude Opus audit",
   "date": "2026-10-19",
   "date_text": "10-19",
   "source_public": null,
   "caveat": "held out of pushed stack AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-10"
   ]
  },
  {
   "id": "F-1404",
   "kind": "fact",
   "claim": "SMALLCLS: 1.7B hazard naming on fresh emergencies",
   "value": "right 14/45 (precision 37 %); 4B 32/45 (74 %); hand list 7/45",
   "ci": "n 45",
   "n": "45",
   "seeds": null,
   "set": "fresh agent-written, dev",
   "set_status": "dev",
   "judge": "count",
   "date": "2026-10-19",
   "date_text": "10-19",
   "source_public": null,
   "caveat": "original report overstated \"recall 38/45\"",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-11"
   ]
  },
  {
   "id": "F-1405",
   "kind": "fact",
   "claim": "RELGATE: 4B as source picker",
   "value": "right source 72 % (31/43, G3 PASS); chunk relevance AUC 0.83 and sufficiency 0.72 FAIL",
   "ci": "held-out half B; 150 contexts",
   "n": null,
   "seeds": null,
   "set": "public self-tests, dev",
   "set_status": "dev",
   "judge": "AUC / accuracy",
   "date": "2026-10-19",
   "date_text": "10-19",
   "source_public": null,
   "caveat": "borderline vs 63 % on half A",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-12"
   ]
  },
  {
   "id": "F-1406",
   "kind": "fact",
   "claim": "V-para 4B gap decomposition",
   "value": "gap 0.586 ratio points: wrong 0.309 (23 items), absent 0.239 (24), thin 0.038 (4), refusals 0",
   "ci": "n 51, s42",
   "n": "51",
   "seeds": null,
   "set": "V-para (dev-style since 10-09)",
   "set_status": "dev-style since 2026-10-09",
   "judge": "Claude Opus judge.v1; LOSSMAP labels",
   "date": "2026-10-19",
   "date_text": "10-19",
   "source_public": null,
   "caveat": "single-seed labels, shares ±0.05",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-13"
   ]
  },
  {
   "id": "F-1407",
   "kind": "fact",
   "claim": "Facts for V-para are mostly nowhere",
   "value": "in installed kit 6/51; all packs on disk 10/51; full English Wikipedia 9/34 absent items",
   "ci": "n 51 / 34",
   "n": "51",
   "seeds": null,
   "set": "V-para",
   "set_status": "dev-style since 2026-10-09",
   "judge": "coverage counts",
   "date": "2026-10-19",
   "date_text": "10-19",
   "source_public": null,
   "caveat": "The 34 are the deep tier's absent items; the 4B's absent set is 24, of which 17 are shared.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-14"
   ]
  },
  {
   "id": "F-1408",
   "kind": "fact",
   "claim": "Retrieval kept nothing on 29/51 V-para items, carrying 58 % of the 4B loss",
   "value": "29/51; 58 %",
   "ci": "n 51",
   "n": "51",
   "seeds": null,
   "set": "V-para",
   "set_status": "dev-style since 2026-10-09",
   "judge": "trace",
   "date": "2026-10-19",
   "date_text": "10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-15"
   ]
  },
  {
   "id": "F-1409",
   "kind": "fact",
   "claim": "Abstention would not pay under judge.v1",
   "value": "replacing every wrong item with a decline: -0.053 raw; only confident-wrong: -0.005",
   "ci": "n 51",
   "n": "51",
   "seeds": null,
   "set": "V-para",
   "set_status": "dev-style since 2026-10-09",
   "judge": "recomputed oracle",
   "date": "2026-10-19",
   "date_text": "10-19",
   "source_public": null,
   "caveat": "judge.v1 floor scoring; the new ruler changes this (true zero)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-16"
   ]
  },
  {
   "id": "F-1410",
   "kind": "fact",
   "claim": "Rubric floor share on V-para",
   "value": "0.206 of raw (3·51/744); raw 0.413 = floor-adj 0.261",
   "ci": "n 51",
   "n": "51",
   "seeds": null,
   "set": "V-para",
   "set_status": "dev-style since 2026-10-09",
   "judge": "arithmetic",
   "date": "2026-10-19",
   "date_text": "10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-17"
   ]
  },
  {
   "id": "F-1411",
   "kind": "fact",
   "claim": "Estimated ceiling of all 8 GB levers on V-para",
   "value": "about +0.01-0.02 raw",
   "ci": "estimate, untested",
   "n": null,
   "seeds": null,
   "set": "V-para",
   "set_status": "dev-style since 2026-10-09",
   "judge": "estimate",
   "date": "2026-10-19",
   "date_text": "10-19",
   "source_public": null,
   "caveat": "estimate only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-18"
   ]
  },
  {
   "id": "F-1412",
   "kind": "fact",
   "claim": "One newly covered absent item is worth",
   "value": "~+0.003-0.005 raw on V-para",
   "ci": "arithmetic",
   "n": null,
   "seeds": null,
   "set": "V-para",
   "set_status": "dev-style since 2026-10-09",
   "judge": "estimate",
   "date": "2026-10-19",
   "date_text": "10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-19"
   ]
  },
  {
   "id": "F-1413",
   "kind": "fact",
   "claim": "NIST crypto pack self-test",
   "value": "0.430 vs 0.422 (+0.008)",
   "ci": "CI -0.012..+0.030; 20 items",
   "n": "20",
   "seeds": null,
   "set": "pack self-test, dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-19",
   "date_text": "10-19",
   "source_public": null,
   "caveat": "user-selected pack",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-20"
   ]
  },
  {
   "id": "F-1414",
   "kind": "fact",
   "claim": "Visa pack self-test",
   "value": "0.366 vs 0.359 (+0.007)",
   "ci": "CI -0.024..+0.041; 20 items",
   "n": "20",
   "seeds": null,
   "set": "pack self-test, dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": "kept on 1/20 (Wikivoyage wins)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-21"
   ]
  },
  {
   "id": "F-1415",
   "kind": "fact",
   "claim": "v0.4 ruler item set frozen: 75 items (71 headline + 4 anchors); checklist points supported on the open web by at least one family and contradicted by none (the 32 balance-round points by both families)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.4 (frozen; never for engine work)",
   "set_status": "frozen ruler set (held-out for the ruler; counts only)",
   "judge": null,
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": "items written by the two judge families (spec states deviation) | Verification note: says '329 points all supported'; only the 32 balance-round points are stated as supported by both families. The earlier 302/307 rule was 'supported by at least one family and contradicted by none' . Honest: 329 points supported on the open web (at least one family, none contradicted; new 32 by both). Numbers withheld: this row did not pass the verify pass as stated (OVERCLAIM); see the confirmed rows on the same result.",
   "verification": "OVERCLAIM",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-1820-22"
   ]
  },
  {
   "id": "F-1416",
   "kind": "fact",
   "claim": "v0.4 candidate writers",
   "value": "Claude 36, OpenAI gpt-5.6-sol 34; 302/307 points supported, 5 disputed",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.4 drafts",
   "set_status": "frozen ruler set (held-out for the ruler; counts only)",
   "judge": null,
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-1820-23"
   ]
  },
  {
   "id": "F-1417",
   "kind": "fact",
   "claim": "LITE17: prompt levers for the 1.7B",
   "value": "thinking@256 -0.002, quote-then-answer +0.002, chunk drop +0.001; 1.7B ~0.50 vs 4B ~0.61",
   "ci": "150 pack items",
   "n": null,
   "seeds": null,
   "set": "pack self-tests, dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": "gap broad across packs (+0.06 to +0.19)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-24"
   ]
  },
  {
   "id": "F-1418",
   "kind": "fact",
   "claim": "Pixel 8a speed, 1.7B vs 4B",
   "value": "1.7B 13-16 tok/s, answers 13-16 s; 4B 3.6-6 tok/s, 29-46 s; 4B LMK kill at 4.4 GB RSS",
   "ci": "9-question runs",
   "n": null,
   "seeds": null,
   "set": "public dev questions",
   "set_status": "dev",
   "judge": "timing",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": "4B 73/73 only after a reboot | VERIFIER NOTE: Pre-BATCHRET, 9 public dev questions; not comparable with the 73-question p50 of 10-21.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-25"
   ]
  },
  {
   "id": "F-1419",
   "kind": "fact",
   "claim": "Cost of the pivot",
   "value": "1.7B reading gap vs 4B -0.112 with packs (LITE17)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "pack self-tests, dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-26"
   ]
  },
  {
   "id": "F-1420",
   "kind": "fact",
   "claim": "LITEDECLINE: compact-tier declines",
   "value": "declines 94/237 when ON; releasing all five: U0 +0.170 (+0.141..+0.199), confident-error count 17 -> 34 (rate +0.025, CI -0.013..+0.063)",
   "ci": "v0.1 57 + 180 self-tests, seed 42",
   "n": null,
   "seeds": "1",
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku; U0 = ratio with declines scored 0; CE = non-declined answers with correctness <= 1.5",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-27"
   ]
  },
  {
   "id": "F-1421",
   "kind": "fact",
   "claim": "LITELAT: phone time to first token breakdown, 1.7B",
   "value": "retrieval 57 % (~435 serial SQL calls), prefill 36 %; TTFT p50 9.4 s",
   "ci": "7 runs x 9 items",
   "n": "9",
   "seeds": null,
   "set": "litelat-9 public dev",
   "set_status": "dev",
   "judge": "timing",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-28"
   ]
  },
  {
   "id": "F-1422",
   "kind": "fact",
   "claim": "BATCHRET",
   "value": "SQL calls per retrieve 937 -> 48; retrieval byte-identical 390/390",
   "ci": "390 items",
   "n": "390",
   "seeds": null,
   "set": "public dev",
   "set_status": "dev",
   "judge": "parity",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-29"
   ]
  },
  {
   "id": "F-1423",
   "kind": "fact",
   "claim": "BATCHRET on the Pixel 8a, 1.7B",
   "value": "TTFT 9.12 -> 5.55 s (-39 %); retrieval 5.04 -> 1.45 s; total 12.39 -> 9.40 s; safety card 8.27 -> 3.10 s",
   "ci": "9 items, A B A B, thermal 0",
   "n": "9",
   "seeds": null,
   "set": "litelat-9",
   "set_status": "unspecified",
   "judge": "timing",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-30"
   ]
  },
  {
   "id": "F-1424",
   "kind": "fact",
   "claim": "CMP30OFF on Pixel 8a (declines released + warning)",
   "value": "1.7B U0 0.394 vs Qwen2.5-1.5B 0.349 (+0.045, CI -0.009..+0.100); correctness +0.46 (+0.02..+0.90); decode 17.6 vs 11.7 tok/s (1.5B faster)",
   "ci": "n 26 judged",
   "n": "26",
   "seeds": null,
   "set": "v0.1 public dev",
   "set_status": "dev",
   "judge": "Claude Haiku 5.5-low",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": "half of warned answers still confident errors | Verification note: Speed mix: 18.6 vs 12.9 tok/s is the CMP30 run ; in CMP30OFF itself decode was 17.6 vs 11.7 tok/s .",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-31"
   ]
  },
  {
   "id": "F-1425",
   "kind": "fact",
   "claim": "LITEPACKS MAIN",
   "value": "no pack set gains U >= +0.03; deltas -0.004..-0.012; K0 U 0.406",
   "ci": "n 97, seed 42",
   "n": "97",
   "seeds": "1",
   "set": "MAIN 97 (v0.1 57 + cooking-se 20 + money-se 20), dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-32"
   ]
  },
  {
   "id": "F-1426",
   "kind": "fact",
   "claim": "LITEPACKS pack-matched",
   "value": "K1 +0.036, K2 +0.045, K3 +0.069 (CI low > 0)",
   "ci": "80 self-tests",
   "n": null,
   "seeds": null,
   "set": "pack self-tests, dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": "context ~2x tokens; +0.74-2.5 GB disk",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-33"
   ]
  },
  {
   "id": "F-1427",
   "kind": "fact",
   "claim": "CE taxonomy of 1.7B confident errors",
   "value": "20/27 had no relevant source; 6 reading failures; 15 fixable by new pack content",
   "ci": "27 items",
   "n": "27",
   "seeds": null,
   "set": "K0 problems, dev",
   "set_status": "dev",
   "judge": "Claude Haiku labels (leads)",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": "labels are leads",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-1820-34"
   ]
  },
  {
   "id": "F-1428",
   "kind": "fact",
   "claim": "Token confidence cannot gate errors",
   "value": "logprob AUC 0.530 (0.400-0.657); first-token entropy AUC 0.488",
   "ci": "n 97",
   "n": "97",
   "seeds": null,
   "set": "MAIN 97, dev",
   "set_status": "dev",
   "judge": "AUC",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-35"
   ]
  },
  {
   "id": "F-1429",
   "kind": "fact",
   "claim": "Groundedness LoRA gate (support)",
   "value": "AUC 0.965; 48/57 1.7B source-present answers unsupported",
   "ci": "n 57",
   "n": "57",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku support labels, kappa 0.79",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-36"
   ]
  },
  {
   "id": "F-1430",
   "kind": "fact",
   "claim": "REFUSE-UNSUPPORTED",
   "value": "visible confident errors 7 -> 3 of 177 under Claude Opus (re-judge of the 38 deciding answers; Claude Haiku 10 -> 4); false refusals 0/15; U on answered +0.030 (+0.005..+0.057, Haiku); coverage 0.435 -> 0.328; +9.3 s p50 projected on the Pixel 8a",
   "ci": "n 177",
   "n": "177",
   "seeds": null,
   "set": "MAIN 97 + 80 self-tests, dev",
   "set_status": "dev",
   "judge": "Claude Haiku 5.5-low; Claude Opus re-judge for the confident-error count",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": "GATELAT: no cheaper variant keeps AUC | Verification note: Judge-dependence: Haiku 10 -> 4 of 177; Opus re-judge of the 38 deciding answers 7 -> 3 of 177 ; the claim text should quote 7 -> 3 with Opus named. +9.3 s is projected.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-37"
   ]
  },
  {
   "id": "F-1431",
   "kind": "fact",
   "claim": "Judge v2 (binary checklist) test-retest",
   "value": "CE test-retest 0.966 vs v1 0.924 (needed 0.974); SD 0.80x (needed 0.60x) -> NO-GO",
   "ci": "frozen rule",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": null,
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-38"
   ]
  },
  {
   "id": "F-1432",
   "kind": "fact",
   "claim": "Haiku vs Opus on confident-error labels",
   "value": "agree 39/47 (kappa 0.63); Haiku over-calls; Opus +0.106 more lenient on correctness",
   "ci": "n 47",
   "n": "47",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": null,
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-39"
   ]
  },
  {
   "id": "F-1433",
   "kind": "fact",
   "claim": "Null calibration of screens",
   "value": "a whitespace-only patch or a new seed fixes 23-31 % of problem items",
   "ci": "4 null runs",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "net fixes",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-40"
   ]
  },
  {
   "id": "F-1434",
   "kind": "fact",
   "claim": "Gemma 4 E2B on MAIN 97",
   "value": "Claude Haiku: dU +0.028 (+0.006..+0.052), stage 2 FAIL by 0.002; two-seed pooled +0.029 (+0.009..+0.048). Claude Opus (seed 42): +0.031 (+0.005..+0.058), GO by 0.001",
   "ci": "n 97",
   "n": "97",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku 5.5-low and Claude Opus (judge-dependent verdict)",
   "date": "2026-10-20",
   "date_text": "10-20/21",
   "source_public": null,
   "caveat": "gain mostly fewer refusals | Verification note: Judge-dependent: under Opus the same seed-42 pair is dU +0.031 (+0.005..+0.058), GO by 0.001 ; the Haiku FAIL stands under Haiku. Gain mostly from fewer refusals.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-41"
   ]
  },
  {
   "id": "F-1435",
   "kind": "fact",
   "claim": "Gemma 4 E2B on the Pixel 8a",
   "value": "73/73, answer p50 11.8 s, p90 24.3 s, PSS 4.52 GB; raw 0.452 vs phone 4B 0.535 (-0.083, -0.123..-0.047)",
   "ci": "57 ids",
   "n": "57",
   "seeds": null,
   "set": "v0.1 public dev",
   "set_status": "dev",
   "judge": "Claude Haiku 5.5-low raw ratio",
   "date": "2026-10-21",
   "date_text": "10-21",
   "source_public": null,
   "caveat": "measured in our bench app, not the submitted app",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-42"
   ]
  },
  {
   "id": "F-1436",
   "kind": "fact",
   "claim": "Ministral-3-3B + citation prompt v2 stage 2",
   "value": "U +0.062 (+0.037..+0.089) vs 1.7B",
   "ci": "n 97, 1 seed",
   "n": "97",
   "seeds": "1",
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "date_text": "10-21",
   "source_public": null,
   "caveat": "~3x phone time (projected p50 21.6-25.6 s). Superseded in the same week: under Claude Opus the PASS stands (+0.047, +0.021..+0.075) but the line was closed as dominated by the 4B (-0.037 Opus, 61 items) and not proposed for the app.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-43"
   ]
  },
  {
   "id": "F-1437",
   "kind": "fact",
   "claim": "Device parity of the bench app on Pixel 8a 4B",
   "value": "phone - desktop +0.006 (-0.015..+0.030)",
   "ci": "57 ids",
   "n": "57",
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "date_text": "10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-44"
   ]
  },
  {
   "id": "F-1438",
   "kind": "fact",
   "claim": "Facts table r9",
   "value": "driving side 197, calling code 196, mains Hz 191; audit 0 wrong; FP 0/675",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "public",
   "set_status": "public dev",
   "judge": "audit",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-45"
   ]
  },
  {
   "id": "F-1439",
   "kind": "fact",
   "claim": "Persona map union size",
   "value": "A∪ 6 text packs 435 MB; ~498 MB with gazetteer, default kit, embedder",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "n/a",
   "judge": "bytes",
   "date": "2026-10-20",
   "date_text": "10-20",
   "source_public": null,
   "caveat": "not frozen",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-1820-46"
   ]
  },
  {
   "id": "F-1440",
   "kind": "fact",
   "claim": "Standing bar numbers unchanged by this slice",
   "value": "V-para floor-adj 4B 0.261 / deep+think 0.436 (Claude); cross-family 0.361 / 0.577; win-rate ~0",
   "ci": "2 seeds",
   "n": null,
   "seeds": "2",
   "set": "V-para",
   "set_status": "dev-style since 2026-10-09",
   "judge": "Claude Opus / OpenAI gpt-5.6-sol post-hoc",
   "date": "2026-10-19",
   "date_text": "10-19",
   "source_public": null,
   "caveat": "restatement, no new run",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-1820-47"
   ]
  },
  {
   "id": "F-1441",
   "kind": "null-result",
   "claim": "Succinct-honest prompt (S1): 49 capability turns",
   "value": "refusals 2 -> 7/49",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: fails refusal gate",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-1820-01"
   ]
  },
  {
   "id": "F-1442",
   "kind": "null-result",
   "claim": "Capability sentence (D3/D4) to make Detailed write code: 21 code asks",
   "value": "1-2/21 code",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: route code asks instead",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-1820-02"
   ]
  },
  {
   "id": "F-1443",
   "kind": "null-result",
   "claim": "LFM2.5-8B-A1B as model: audits, PACKREAD, SAFEPACK oracle",
   "value": "more wrong claims, 11-13 unsafe",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: keep Qwen AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-1820-03"
   ]
  },
  {
   "id": "F-1444",
   "kind": "null-result",
   "claim": "PROMPT3M (honesty prompts on 3 small models, no source): partial",
   "value": "not completed",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: overlapped PACKREAD; direction review closed it",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-1820-04"
   ]
  },
  {
   "id": "F-1445",
   "kind": "null-result",
   "claim": "SAFEROUTE (emergency retrieval): tuned and scored on the same 30 items; expected not to carry to new wording (a hand hazard list fired on 7/45 fresh emergencies); held out of the shipped stack",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: held from the PR | Verification note: 'out-of-sample reach small' is the reviewer's inference from the hand hazard list firing on 7/45 fresh emergencies, not a measured saferoute out-of-sample run . Honest: 'expected not to carry to new wording (hand list 7/45 on fresh items); in-sample, tuned and scored on the same 30'. Numbers withheld: this row did not pass the verify pass as stated (OVERCLAIM); see the confirmed rows on the same result.",
   "verification": "OVERCLAIM",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-1820-05"
   ]
  },
  {
   "id": "F-1446",
   "kind": "null-result",
   "claim": "1.7B hazard classifier to drive retrieval: 45 fresh",
   "value": "37 % precision",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: unsafe to feed retrieval AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-1820-06"
   ]
  },
  {
   "id": "F-1447",
   "kind": "null-result",
   "claim": "Relevance / sufficiency gates (RELGATE G1/G2): 642 chunks",
   "value": "AUC 0.83 / 0.72",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: below frozen bar",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-1820-07"
   ]
  },
  {
   "id": "F-1448",
   "kind": "null-result",
   "claim": "CHAINWHEN (4B source pick before answer): 86 triggered",
   "value": "+0.019 (-0.006..+0.044)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: NO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-1820-08"
   ]
  },
  {
   "id": "F-1449",
   "kind": "null-result",
   "claim": "New packs for V-para (NIST, visa): self-tests",
   "value": "+0.008, +0.007",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: inert; facts mostly absent everywhere",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-1820-09"
   ]
  },
  {
   "id": "F-1450",
   "kind": "null-result",
   "claim": "LITE17 prompt tricks for 1.7B: 150 items",
   "value": "<= ±0.002",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: gap is model reading",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-1820-10"
   ]
  },
  {
   "id": "F-1451",
   "kind": "null-result",
   "claim": "LITEPACKS general packs default: MAIN 97",
   "value": "-0.004..-0.012",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: NO CHANGE",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-1820-11"
   ]
  },
  {
   "id": "F-1452",
   "kind": "null-result",
   "claim": "Honest prompt / short context / quote-first / greedy / vendor sampling on 1.7B: stage-1 screens",
   "value": "fixed 1-4 of 21-23 CE",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: below bar (later shown bar was at null)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-1820-12"
   ]
  },
  {
   "id": "F-1453",
   "kind": "null-result",
   "claim": "Confidence-based gating (logprob, first-token entropy): 97 items",
   "value": "AUC ~0.49-0.53",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: no signal",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-1820-13"
   ]
  },
  {
   "id": "F-1454",
   "kind": "null-result",
   "claim": "OCC-RAG 1.7B / 0.6B: screens",
   "value": "fixes by refusing; format breaks",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: line closed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-1820-14"
   ]
  },
  {
   "id": "F-1455",
   "kind": "null-result",
   "claim": "GATE17 on correctness: 57",
   "value": "AUC 0.496",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: wrong target; re-run on support PASSED",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-1820-15"
   ]
  },
  {
   "id": "F-1456",
   "kind": "null-result",
   "claim": "Retrieval hygiene V1orC: K1 screen",
   "value": "net 2 (bar 9)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: not a quality lever",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-1820-16"
   ]
  },
  {
   "id": "F-1457",
   "kind": "null-result",
   "claim": "Cheaper groundedness gate (GATELAT): 124",
   "value": "AUC 0.934 -> 0.66 at 1.8 s",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: no cheap variant",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-1820-17"
   ]
  },
  {
   "id": "F-1458",
   "kind": "null-result",
   "claim": "Judge v2 binary checklist: test-retest",
   "value": "below frozen rule",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: keep v1 for now",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-1820-18"
   ]
  },
  {
   "id": "F-1459",
   "kind": "null-result",
   "claim": "Q8_0 quant of 1.7B: screens",
   "value": "null band; x0.66 decode",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-1820-19"
   ]
  },
  {
   "id": "F-1460",
   "kind": "null-result",
   "claim": "Qwen3.5-2B, SmolLM3-3B, Jan-v1-edge, FOTAPOL LoRA: screens",
   "value": "citation drops -> false refusals",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-1820-20"
   ]
  },
  {
   "id": "F-1461",
   "kind": "null-result",
   "claim": "granite-4.0-1b: stage 2",
   "value": "dU +0.0003",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: equal to 1.7B",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-1820-21"
   ]
  },
  {
   "id": "F-1462",
   "kind": "null-result",
   "claim": "Shorter context, second try (SHORTCTX2): GATE v2 K1",
   "value": "net 8 (bar 9), 6/20 false refusals",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-1820-22"
   ]
  },
  {
   "id": "F-1463",
   "kind": "null-result",
   "claim": "bench v0.3 (non-Claude held-out): not built",
   "value": "—",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: D5 timing; reframed into v0.4 ruler",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "N-u-1820-23"
   ]
  },
  {
   "id": "F-1464",
   "kind": "null-result",
   "claim": "Safety cards before retrieval (C): not built",
   "value": "—",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: archived; BATCHRET already cut card wait 8.3 -> 3.1 s",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-1820-24"
   ]
  },
  {
   "id": "F-1465",
   "kind": "fact",
   "claim": "Earlier: Qwen3-4B is the claim/fast-tier engine for 8 GB phones. Changed by 1.7B pivot: main 8 GB engine = Qwen3-1.7B; 4B = optional \"more complete\" tier for 10-12 GB phones",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "S-u-1820-01"
   ]
  },
  {
   "id": "F-1466",
   "kind": "fact",
   "claim": "Earlier: Lite line (1.7B) closed: 1.7B + X2b = 4B only on pack-matched dev self-tests; claim kit -0.111. Changed by 1.7B pivot: the lite gap is accepted as the price of speed; the line reopens as LITE-MAIN",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "S-u-1820-02"
   ]
  },
  {
   "id": "F-1467",
   "kind": "fact",
   "claim": "Earlier: Bench v0.3 (non-Claude, ~60-110 items) is the path to a clean held-out number. Changed by D5 (10-18) then v0.4 ruler (10-20): v0.3 dropped; a 75-item v0.4 written by Claude + OpenAI gpt-5.6-sol (two judge families, a stated deviation) frozen as the ruler set",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "S-u-1820-03"
   ]
  },
  {
   "id": "F-1468",
   "kind": "fact",
   "claim": "Earlier: Headline metric: raw and floor-adjusted ratio vs a 250-word Claude Opus reference, judge.v1 1-5 rubric, win-rate first",
   "value": "Changed by new ruler: headline = U-total (true zero; refusal = 0) with coverage split + task-success ratio; uncapped two-family web references; min over two judge families; old numbers become dev evidence",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "No number on the new ruler exists by 2026-10-20; the judge.v1 numbers remain the only standing numbers of that week. Metric: ruler v0.4 U-total (ratio of sums against the reference mass, 71 headline items, health weight 0.5); never compare with the judge.v1 ratio.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "S-u-1820-04"
   ]
  },
  {
   "id": "F-1469",
   "kind": "fact",
   "claim": "Earlier: Scored system = desktop TS engine (cut 7). Changed by new ruler: scored system = submitted APK on a named <= 12 GB phone; desktop = parity evidence",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "S-u-1820-05"
   ]
  },
  {
   "id": "F-1470",
   "kind": "fact",
   "claim": "Earlier: Abstention does not pay (decline scores the floor). Changed by ruler true-zero + refuse-by-default product: refusal now scores 0 but confident errors are counted separately; product refuses by default when no source is kept",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "S-u-1820-06"
   ]
  },
  {
   "id": "F-1471",
   "kind": "fact",
   "claim": "Earlier: Compact-model behaviour on no source: decline (engine) / answer under warning (10-20 morning proposal). Changed by 20:24 author decision: refuse by default + \"show anyway\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "S-u-1820-07"
   ]
  },
  {
   "id": "F-1472",
   "kind": "fact",
   "claim": "Earlier: Submission deadline 2026-10-20. Changed by later information: public decision date 2026-10-31; a non-public working date is used internally (never cite)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "S-u-1820-08"
   ]
  },
  {
   "id": "F-1473",
   "kind": "fact",
   "claim": "Earlier: Detailed persona prompt = maintainer's original. Changed by HONESTPROMPT: D4 honest + capability rule recommended",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "S-u-1820-09"
   ]
  },
  {
   "id": "F-1474",
   "kind": "fact",
   "claim": "Earlier: Screen bar \"6 of the control's problems fixed\". Changed by GATE v2: bar = measured null + 2 (nulls fix 23-31 %); earlier stage-1 NO-GOs were unresolvable",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "S-u-1820-10"
   ]
  },
  {
   "id": "F-1475",
   "kind": "fact",
   "claim": "Earlier: SAFEPACK \"unsafe 0 -> 3\". Changed by direction review: 2 caused by patch + 1 audit split; 2 tornado regressions",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "S-u-1820-11"
   ]
  },
  {
   "id": "F-1476",
   "kind": "fact",
   "claim": "Earlier: SMALLCLS \"1.7B recall 38/45\". Changed by direction review: right hazard 14/45",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "S-u-1820-12"
   ]
  },
  {
   "id": "F-1477",
   "kind": "fact",
   "claim": "Earlier: Site upload: preprint v0.1 possible before closure. Changed by D7: only v1.0 at closure",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "S-u-1820-13"
   ]
  },
  {
   "id": "F-1478",
   "kind": "event",
   "claim": "Gemma 4 mobile deep research (10-agent workflow, 92 findings, 35 confirmed, 10 corrected): opt-in entry with thinking guards; 1.7B stays default",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-2122-01"
   ]
  },
  {
   "id": "F-1479",
   "kind": "event",
   "claim": "GEMMA-VS-4B on MAIN 97 (Haiku): Gemma captures 38 % of the 4B gain",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-2122-02"
   ]
  },
  {
   "id": "F-1480",
   "kind": "event",
   "claim": "Direction proposal reviewed by a 9-agent adversarial workflow: 71 findings stand, 5 fall, no wrong number",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-2122-03"
   ]
  },
  {
   "id": "F-1481",
   "kind": "event",
   "claim": "Pixel 8a: Qwen3-1.7B on 73 questions p50 4.2 s vs Gemma 11.8 s",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-2122-04"
   ]
  },
  {
   "id": "F-1482",
   "kind": "event",
   "claim": "HON-1/HON-2: warning line precision 0.39; \"show anyway\" button gives ~7 correct vs 35 CE per 100 reveals (liability as specified)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-2122-05"
   ]
  },
  {
   "id": "F-1483",
   "kind": "event",
   "claim": "A-union (6 packs + gazetteer) on Pixel 8a with 1.7B / Gemma / 4B: no memory cost, +2.4 s serial pack search, 0 LMK",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "date_text": "2026-10-21 03:51-04:35",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-2122-06"
   ]
  },
  {
   "id": "F-1484",
   "kind": "event",
   "claim": "Competitor refresh: a competitor team ships a 35B streamed model + full Wikipedia; their 65 % is on 61 maintainer questions, not comparable",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "date_text": "2026-10-21 04:22-04:27",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "T-u-2122-07"
   ]
  },
  {
   "id": "F-1485",
   "kind": "event",
   "claim": "RV2 separability ladder (Haiku): ruler separates 4B from 1.7B/Gemma; 1.7B level with closed-book 1.5B",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-2122-08"
   ]
  },
  {
   "id": "F-1486",
   "kind": "event",
   "claim": "CV5 ladder under Opus: 1.7B vs closed-book 1.5B becomes separable (judge-dependent sentence)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-2122-09"
   ]
  },
  {
   "id": "F-1487",
   "kind": "event",
   "claim": "GEMMA-OPUS-MAIN: Gemma − 1.7B +0.031 s42 under Opus, GO by 0.001",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-2122-10"
   ]
  },
  {
   "id": "F-1488",
   "kind": "event",
   "claim": "Morning report with 11 author decisions published (PT-BR)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-2122-11"
   ]
  },
  {
   "id": "F-1489",
   "kind": "event",
   "claim": "RULING R1: judge = judge.v1 + DECLINE2STAGE zero per detected non-answer",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-2122-12"
   ]
  },
  {
   "id": "F-1490",
   "kind": "event",
   "claim": "GOAL-2026-10-21 set: raise U_8GB +0.03 (Opus) with guards; cut T_8GB 7.6 → 5.0 s",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-2122-13"
   ]
  },
  {
   "id": "F-1491",
   "kind": "event",
   "claim": "GRILL session: honest 8 GB product + open ruler; no 35B delivery; research site; final score on the native APK",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-2122-14"
   ]
  },
  {
   "id": "F-1492",
   "kind": "event",
   "claim": "1.7B engine lines closed one by one at nulls (X2b, routing, keep, coverage reasons, first-aid cards, keep-E)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "date_text": "2026-10-21 13:18-14:40",
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-2122-15"
   ]
  },
  {
   "id": "F-1493",
   "kind": "event",
   "claim": "Simplewiki pack r2 (converter fix) published to the public pack dataset",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-16"
   ]
  },
  {
   "id": "F-1494",
   "kind": "event",
   "claim": "Author: the submission is the maintainer's native Kotlin app (E1 correction)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "T-u-2122-17"
   ]
  },
  {
   "id": "F-1496",
   "kind": "event",
   "claim": "A1-SUCCINCT17: Succinct as small-model default saves ~24 s p50 (projection)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "T-u-2122-19"
   ]
  },
  {
   "id": "F-1498",
   "kind": "event",
   "claim": "E1-PAR on Pixel 8a (bench engine): Promise.all pack search pays (answer 10.32 → 7.39 s)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-21"
   ]
  },
  {
   "id": "F-1499",
   "kind": "event",
   "claim": "G1-SWA5: Gemma multi-turn KV bug hypothesis killed on the phone",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "T-u-2122-22"
   ]
  },
  {
   "id": "F-1500",
   "kind": "event",
   "claim": "RC1-PACKS: no offline route to install packs in RC1",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "T-u-2122-23"
   ]
  },
  {
   "id": "F-1502",
   "kind": "event",
   "claim": "SHIP15B-PACKS no gain; STYLE15B Succinct non-inferior (Haiku)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-25"
   ]
  },
  {
   "id": "F-1505",
   "kind": "event",
   "claim": "Day close: engine rules exhausted; submission = RC1 1.5B Detailed no packs; Succinct is the lever",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "T-u-2122-28"
   ]
  },
  {
   "id": "F-1506",
   "kind": "event",
   "claim": "NOREPACK97: --no-repack makes two hosts byte-identical 97/97",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-29"
   ]
  },
  {
   "id": "F-1507",
   "kind": "event",
   "claim": "Author ruler decisions: verdict on the app with packs, OpenAI gpt-5.6-sol second judge, headline waits for new RC",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-2122-30"
   ]
  },
  {
   "id": "F-1508",
   "kind": "event",
   "claim": "RULER-SPEC v1 r3 frozen (private)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-2122-31"
   ]
  },
  {
   "id": "F-1509",
   "kind": "event",
   "claim": "REFS-V2 Claude half frozen: 75/75 Opus+web references on v0.4 before any app answer",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-2122-32"
   ]
  },
  {
   "id": "F-1510",
   "kind": "event",
   "claim": "RULER-SPEC r4 frozen + FACT-AVAIL labels (19/71 AVAILABLE)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "T-u-2122-33"
   ]
  },
  {
   "id": "F-1511",
   "kind": "event",
   "claim": "DECODE-GROUND closed: decoder-level grounding on the 1.7B at the null",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-2122-34"
   ]
  },
  {
   "id": "F-1512",
   "kind": "event",
   "claim": "QFORM: question form decides retrieval; reference kept 0.88 vs not kept 0.19",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "date_text": "2026-10-21 22:38-22:58",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-35"
   ]
  },
  {
   "id": "F-1513",
   "kind": "event",
   "claim": "Local clean build of the native app with Succinct default (never pushed)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "T-u-2122-36"
   ]
  },
  {
   "id": "F-1514",
   "kind": "event",
   "claim": "RETRIEVE-SWEEP + chunk vectors: delivery up on pack questions, real-item recovery 3/25 (fail by rule)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "date_text": "2026-10-22 00:44-03:41",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-37"
   ]
  },
  {
   "id": "F-1515",
   "kind": "event",
   "claim": "KB snapshot + packs-in-catalog in both clean apps (Android, iOS): setup ~2 min vs ~280 s",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "date_text": "2026-10-22 03:xx",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-38"
   ]
  },
  {
   "id": "F-1516",
   "kind": "event",
   "claim": "Author direction: packs are the strategy; device day stopped",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-39"
   ]
  },
  {
   "id": "F-1517",
   "kind": "event",
   "claim": "Device day report: only the 4B speed row ran (76 s p50, 4.77 tok/s)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "T-u-2122-40"
   ]
  },
  {
   "id": "F-1518",
   "kind": "event",
   "claim": "Author decision: small focused packs, not giant ones",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-41"
   ]
  },
  {
   "id": "F-1519",
   "kind": "event",
   "claim": "T1 search-term extraction fails; author picks composer design \"one button + chips for what changed\"",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "date_text": "2026-10-22 17:xx",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-42"
   ]
  },
  {
   "id": "F-1520",
   "kind": "event",
   "claim": "H4 global-IDF merge: first no-harm pack-composition gain (+6.2 pts kept)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-43"
   ]
  },
  {
   "id": "F-1521",
   "kind": "event",
   "claim": "COMPOSE: 10-13 small packs compose without crowding; answers flat",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "date_text": "2026-10-22 18:30-18:52",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-44"
   ]
  },
  {
   "id": "F-1522",
   "kind": "event",
   "claim": "SEFIX-ALL (search fixes incl. the F3 health fallback): pooled two-seed gain about +0.022/+0.023 with CI low at about 0 (Claude Haiku); rests on F3, withdrawn for safety; the safe set is +0.013 (-0.008..+0.035); claim set (Claude Opus) neutral",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "date_text": "2026-10-22 21:45-22:17",
   "source_public": null,
   "caveat": "Verification note: Pooled two-seed CI is reported three ways: +0.001..+0.048 , -0.001..+0.048 (same arm, COMBINED report), -0.002..+0.047 (+0.022, , on SEFIX-124's basis). Honest: +0.022/+0.023 with CI low at about 0 (touches or straddles 0); and the gain rests on F3, withdrawn for safety . AI-reviewed, not clinically reviewed. Numbers withheld: this row did not pass the verify pass as stated (CONFLICT); see the confirmed rows on the same result.",
   "verification": "CONFLICT",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-45"
   ]
  },
  {
   "id": "F-1523",
   "kind": "event",
   "claim": "Opus on covered domain: where the fact arrives U 0.16 → 0.33",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-46"
   ]
  },
  {
   "id": "F-1524",
   "kind": "event",
   "claim": "F3 health fallback fails safety (unsafe 9 → 13); withdrawn",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-47"
   ]
  },
  {
   "id": "F-1525",
   "kind": "event",
   "claim": "SEFIX-124 (safe set) +0.013: not a lever",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-48"
   ]
  },
  {
   "id": "F-1526",
   "kind": "event",
   "claim": "F3b and detector retrain both fail",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "date_text": "2026-10-22 23:16-23:39",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "T-u-2122-49"
   ]
  },
  {
   "id": "F-1527",
   "kind": "decision",
   "claim": "Qwen3-1.7B stays the 8 GB recommendation \"on a latency and memory constraint, not on quality\"; model swaps below 4B parked until 10-31",
   "value": "stood; superseded in practice by RC1 shipping 1.5B",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "date_text": "2026-10-21 04:xx",
   "source_public": null,
   "caveat": "who decided: delegated coordinator (proposal, reviewed); why: Gemma +0.029 at 1.5× time +1.4 GB; Ministral +0.062 at ~3× time; 4B +0.076 at ~2.5× time; alternatives: Gemma default; 4B default; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-2122-01"
   ]
  },
  {
   "id": "F-1528",
   "kind": "decision",
   "claim": "Gemma 4 E2B only as opt-in catalog entry with thinking guards (thinking off, budget 0, n_predict cap, post-filter); exact file pinned",
   "value": "proposal; Gemma not in RC1 catalog so moot on RC1",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who decided: delegated coordinator; why: phone quality = 1.7B, 1.5-2.8× time, +1.1-1.4 GB; runtime risks (thinking default on, KV growth); alternatives: Gemma default; drop Gemma; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-2122-02"
   ]
  },
  {
   "id": "F-1529",
   "kind": "decision",
   "claim": "RULING R1: judge = judge.v1 unchanged + DECLINE2STAGE zero per detected non-answer; D 0.058 kept as sensitivity row",
   "value": "adopted into ruler r3/r4",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who decided: delegated coordinator (override of a frozen FAIL, accepted with conditions by reviewer); why: judge v1.1/v1.1b changed CE verdicts (5-7/40 flips); DECLINE2STAGE's 2 bar-(b) hits were engine refusal strings; alternatives: judge v1.1; decline subtraction only; reversible: yes (pre-exposure)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-2122-03"
   ]
  },
  {
   "id": "F-1530",
   "kind": "decision",
   "claim": "GOAL: two numbers U_8GB (+0.03 Opus, CE-shown must not rise) and T_8GB (7.6 → 5.0 s)",
   "value": "T_8GB re-based to the native app same day; U_8GB lines closed at nulls",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who decided: author; why: focus the last days on measurable engine gains; alternatives: open exploration; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-2122-04"
   ]
  },
  {
   "id": "F-1531",
   "kind": "decision",
   "claim": "Winning thesis = honest 8 GB product + open ruler",
   "value": "stood until 10-22 pivot reframed it",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who decided: author (GRILL); why: 35B not deliverable on 8 GB in time; alternatives: compete on the deep tier; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-2122-05"
   ]
  },
  {
   "id": "F-1532",
   "kind": "decision",
   "claim": "35B deep mode not delivered; deep tier appears only as roadmap with desktop numbers",
   "value": "stood",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who decided: author (GRILL); why: focus on 8 GB UX; no Android expert-streaming build; alternatives: 12 GB deep tier; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "D-u-2122-06"
   ]
  },
  {
   "id": "F-1533",
   "kind": "decision",
   "claim": "8 GB default model: Qwen3-1.7B, 4B optional (sent to maintainer only within the short package)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who decided: author. Same set and phone: 1.7B p50 4.2 s vs 4B p50 25.5 s over all 73 v0.1 items (all-items p50, including deterministic and declined items). Numbers withheld: this row did not pass the verify pass as stated (OVERCLAIM); see the confirmed rows on the same result.",
   "verification": "OVERCLAIM",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-2122-07"
   ]
  },
  {
   "id": "F-1534",
   "kind": "decision",
   "claim": "No source found → refuse by default + \"show anyway\", shown answers carry a warning line",
   "value": "later narrowed: show-anyway only behind second confirmation, never on health (REVEAL-CENSUS); RC1 has no warning line",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who decided: author (GRILL); why: honesty on 8 GB; alternatives: always answer; reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-2122-08"
   ]
  },
  {
   "id": "F-1535",
   "kind": "decision",
   "claim": "Article = research site on a project subdomain built from the public research repo; story = timeline of decisions, not a score line",
   "value": "stands (this publication)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who decided: author (GRILL); why: full control of look; avoid cross-set score charts; alternatives: pages on existing site; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-2122-09"
   ]
  },
  {
   "id": "F-1536",
   "kind": "decision",
   "claim": "Final score = submitted native APK on Pixel 8a, v0.4 frozen (75 items), Claude + OpenAI gpt-5.6-sol judges, frontier + web references",
   "value": "ruler r3/r4 built for it; run pending a new RC",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who decided: author (GRILL); why: the bounty judges the app; alternatives: engine-only score; reversible: partly (depends on RC)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-2122-10"
   ]
  },
  {
   "id": "F-1537",
   "kind": "decision",
   "claim": "Submission = the maintainer's native Kotlin app, not the RN bench engine",
   "value": "re-based speed lines (bench levers page-only) and claim section 0",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who decided: author; why: that is what will be submitted; alternatives: engine-only claim; reversible: no (fact of the submission)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-2122-11"
   ]
  },
  {
   "id": "F-1538",
   "kind": "decision",
   "claim": "No 512-token cap ask (CAP512 not a lever)",
   "value": "dropped",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who decided: delegated coordinator (frozen rule); why: saves 0 s p50, cuts 10.9 % mid-sentence; alternatives: cap ask in NOTE-1; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-2122-12"
   ]
  },
  {
   "id": "F-1541",
   "kind": "decision",
   "claim": "Verdict arm = app with packs (A-union); OpenAI gpt-5.6-sol as second judge family; headline waits for new RC; ruler in its own session",
   "value": "in r3/r4",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who decided: author; why: thesis \"app + packs answers well\"; two families; alternatives: verdict on app without packs; reversible: yes (pre-run)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-2122-15"
   ]
  },
  {
   "id": "F-1542",
   "kind": "decision",
   "claim": "Freeze RULER-SPEC v1 r3 (private)",
   "value": "frozen sha 0fac2286",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who decided: delegated coordinator + reviewer ACCEPT WITH EDITS; why: pre-registration before any v0.4 answer; alternatives: keep binary checklist judge (NO-GO earlier); reversible: no (frozen)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-2122-16"
   ]
  },
  {
   "id": "F-1543",
   "kind": "decision",
   "claim": "Freeze r4: add information-available split (FACT-AVAIL) + ORACLE/ORACLE-0 rows; U-total stays verdict",
   "value": "frozen sha 82772e33; labels quarantined",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who decided: author asked; coordinator + reviewer (H1-H8); why: separate \"packs lack it\" from \"app fails to deliver it\"; alternatives: no split; reversible: no (frozen) Metric: ruler v0.4 U-total (ratio of sums against the reference mass, 71 headline items, health weight 0.5); never compare with the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-2122-17"
   ]
  },
  {
   "id": "F-1544",
   "kind": "decision",
   "claim": "Close decoder-level grounding axis on the 1.7B",
   "value": "closed",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who decided: delegated coordinator (frozen gate); why: all arms net 4-6 vs bar 9; alternatives: AdaCAD (unopened); reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-2122-18"
   ]
  },
  {
   "id": "F-1545",
   "kind": "decision",
   "claim": "Carry chunk vectors forward only in no-displacement \"extra\" mode; fusion harmful on real items",
   "value": "G2 full kit then failed real-item recovery (3/25)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "date_text": "2026-10-22 01:10-02:43",
   "source_public": null,
   "caveat": "who decided: delegated coordinator; why: fusion lost 6-8/22 real items; alternatives: fusion; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-2122-19"
   ]
  },
  {
   "id": "F-1546",
   "kind": "decision",
   "claim": "Packs are the strategy: engine finds the context when the pack has it, model answers from it; stop expecting the model to answer alone; device day stopped",
   "value": "two lanes (retrieval, coverage); NOTE-5 first ask",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "who decided: author; why: QFORM 0.88 vs 0.19; alternatives: keep model/engine levers; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-2122-20"
   ]
  },
  {
   "id": "F-1547",
   "kind": "decision",
   "claim": "Small focused packs, not giant ones; full Wikipedia = negative control",
   "value": "composition (merge) becomes the central question",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "who decided: author; why: top100k takes 47-55 % of kept slots for net +2; small Q&A packs do not crowd; competitor full-enwiki search 17-21 s under model load; alternatives: full Wikipedia; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-2122-21"
   ]
  },
  {
   "id": "F-1548",
   "kind": "decision",
   "claim": "Composer UX = one options button + chips only for non-default settings; auto search-term chips dropped",
   "value": "design adopted for the fork",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "date_text": "2026-10-22 evening",
   "source_public": null,
   "caveat": "who decided: author; why: T1 term extraction failed (25 → 23 %); alternatives: 4-chip query bar MVP; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-2122-22"
   ]
  },
  {
   "id": "F-1549",
   "kind": "decision",
   "claim": "Withdraw F3 (health no-source fallback) from the search fixes",
   "value": "shippable set = gidf + F1 + F2 + F4 (+0.013, not a lever)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "who decided: delegated coordinator (frozen safety rule); why: unsafe 9 → 13 on 65 probes; alternatives: keep F3 for +0.010; reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-2122-23"
   ]
  },
  {
   "id": "F-1550",
   "kind": "decision",
   "claim": "Delivery is measured at chunk level from now on",
   "value": "standing method",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "who decided: delegated coordinator; why: article-level overstated F1b by 13 facts; alternatives: article-level; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-2122-24"
   ]
  },
  {
   "id": "F-1551",
   "kind": "decision",
   "claim": "NOTE-2 honesty wording scoped to 1.7B/Gemma; never \"more honest\"; show-anyway behind second confirmation, never on health",
   "value": "drafted",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "date_text": "2026-10-21 03:xx-05:xx",
   "source_public": null,
   "caveat": "who decided: delegated coordinator, author OK pending; why: warning line indeterminate for 4B (precision 0.275); show-anyway 19 correct / 28 CE per 100 (Opus); alternatives: general refuse-by-default; reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-2122-25"
   ]
  },
  {
   "id": "F-1560",
   "kind": "fact",
   "claim": "Packs on the 1.5B model in the research desktop engine (Succinct, K1 packs; not the submitted app, which cannot install packs)",
   "value": "dU +0.015 (−0.036..+0.068); CE-shown 25.5 vs 31.4 per 100; 1.5B+packs − 1.7B K0 −0.082",
   "ci": "n 51, seed 42",
   "n": "51",
   "seeds": "1",
   "set": "CV5 51 (dev)",
   "set_status": "dev",
   "judge": "Claude Opus U (+Claude Haiku +0.011)",
   "date": "2026-10-21",
   "date_text": "10-21 18:20",
   "source_public": null,
   "caveat": "win-rate 0.02 both | Verification note: This is the 1.5B model in OUR desktop engine (Succinct, FACTS off, K1 packs, desktop machine) vs its closed-book run, not RC1 with packs (RC1 cannot install packs).",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "F-u-2122-09"
   ]
  },
  {
   "id": "F-1565",
   "kind": "fact",
   "claim": "Phone, Pixel 8a, 73 questions, bench app",
   "value": "1.7B p50 4.2 s, p90 12.2, 12.4 tok/s, PSS 3.39 GB; Gemma E2B p50 11.8 s, 7.0 tok/s, PSS 4.52 GB",
   "ci": "n 73, 1 run each",
   "n": "73",
   "seeds": "1",
   "set": "v0.1 (dev)",
   "set_status": "dev",
   "judge": "speed",
   "date": "2026-10-21",
   "date_text": "10-21 03:31",
   "source_public": null,
   "caveat": "p50 is over all 73 items including deterministic-layer and declined items: the 1.7B made 38 LLM calls vs Gemma 55 and declines 35 % of v0.1; model-path answer p50 is about 12 s. Not like-for-like with the 9-question 13-16 s runs of 2026-10-20.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-2122-14"
   ]
  },
  {
   "id": "F-1566",
   "kind": "fact",
   "claim": "A-union packs on Pixel 8a: memory and speed",
   "value": "1.7B p50 7.6 vs 4.2 s, PSS 3.41 vs 3.39 GB; Gemma 13.3 vs 11.8 s, PSS 4.38; 4B 19.4 s p50, PSS peak 4.76 GB; 0 LMK",
   "ci": "n 73 each",
   "n": "73",
   "seeds": null,
   "set": "v0.1 (dev)",
   "set_status": "dev",
   "judge": "speed",
   "date": "2026-10-21",
   "date_text": "10-21 03:51-04:35",
   "source_public": null,
   "caveat": "cost = serial pack search (+2.4 s) in the RN bench engine only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-2122-15"
   ]
  },
  {
   "id": "F-1567",
   "kind": "fact",
   "claim": "Bench-engine parallel pack search on Pixel 8a",
   "value": "search 3.89 → 1.93 s; answer 10.32 → 7.39 s",
   "ci": "n 10 per arm",
   "n": "10",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "speed",
   "date": "2026-10-21",
   "date_text": "10-21 17:02",
   "source_public": null,
   "caveat": "bench engine only, page number",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-16"
   ]
  },
  {
   "id": "F-1568",
   "kind": "fact",
   "claim": "Gemma vs 1.7B vs 4B (MAIN 97, Haiku)",
   "value": "U 0.406 / 0.434 / 0.481; 4B − 1.7B +0.076 (+0.050..+0.101); 4B − Gemma +0.048 (+0.021..+0.075); Gemma captures 38 %",
   "ci": "seed 42 (+s43 check)",
   "n": null,
   "seeds": "1",
   "set": "MAIN 97 (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku U",
   "date": "2026-10-21",
   "date_text": "10-21 02:44",
   "source_public": null,
   "caveat": "4B − Gemma under the 0.05 one-seed rule",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-2122-17"
   ]
  },
  {
   "id": "F-1569",
   "kind": "fact",
   "claim": "Gemma vs 1.7B under Opus",
   "value": "U 0.435 → 0.466, dU +0.031 (+0.005..+0.058) s42; s43 +0.041; pooled +0.036 (+0.011..+0.062)",
   "ci": "2 seeds",
   "n": null,
   "seeds": "2",
   "set": "MAIN 97 (dev)",
   "set_status": "dev",
   "judge": "Claude Opus U",
   "date": "2026-10-21",
   "date_text": "10-21 05:55",
   "source_public": null,
   "caveat": "gain mostly from answering more (refusals 46 → 41)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-2122-18"
   ]
  },
  {
   "id": "F-1570",
   "kind": "fact",
   "claim": "CV5 ladder under Opus",
   "value": "1.5B closed-book 0.261, 1.7B K0 0.351, 4B cut 7 phone 0.460; 4B − 1.7B +0.109 (+0.063..+0.160); 1.7B − 1.5B +0.090 (+0.009..+0.170)",
   "ci": "n 51, 1 seed per rung",
   "n": "51",
   "seeds": "1",
   "set": "CV5 51 (dev)",
   "set_status": "dev",
   "judge": "Claude Opus U",
   "date": "2026-10-21",
   "date_text": "10-21 05:28",
   "source_public": null,
   "caveat": "Haiku same frame +0.038 (CI covers 0): judge-dependent | VERIFIER NOTE: Coincidence alert: CV5 closed-book 1.5B 0.261 equals the V-para 4B floor-adjusted 0.261; different sets, models and judges' scopes; never place side by side.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-2122-19"
   ]
  },
  {
   "id": "F-1571",
   "kind": "fact",
   "claim": "Gemma E2B safety behind the warning stack",
   "value": "unsafe 1 vs 1.7B 2 of 65 probes; 0 thinking markers",
   "ci": "n 65",
   "n": "65",
   "seeds": null,
   "set": "safety probes (dev)",
   "set_status": "dev",
   "judge": "reader labels",
   "date": "2026-10-21",
   "date_text": "10-21 04:43",
   "source_public": null,
   "caveat": "our engine, not the maintainer build AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-2122-20"
   ]
  },
  {
   "id": "F-1572",
   "kind": "fact",
   "claim": "Warning line calibration (1.7B)",
   "value": "precision 0.39 (0.32-0.46), recall 0.67",
   "ci": "MAIN",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "date_text": "10-21 03:38",
   "source_public": null,
   "caveat": "Opus re-judge dropped the \"5 % of warned are correct\" sentence",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-2122-21"
   ]
  },
  {
   "id": "F-1573",
   "kind": "fact",
   "claim": "Show-anyway reveals",
   "value": "Opus census: 11 correct vs 16 CE of hidden 1.7B answers (0.69); Haiku: ~7 correct / 35 CE per 100",
   "ci": "MAIN",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Opus / Claude Haiku",
   "date": "2026-10-21",
   "date_text": "10-21",
   "source_public": null,
   "caveat": "only behind second confirmation, never health AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-2122-22"
   ]
  },
  {
   "id": "F-1574",
   "kind": "fact",
   "claim": "Closed-book 1.7B vs with packs",
   "value": "CB 0.337 (0.256-0.425) vs K0 0.351, K0 − CB +0.015 (−0.035..+0.056)",
   "ci": "n 51",
   "n": "51",
   "seeds": null,
   "set": "CV5 51 (dev)",
   "set_status": "dev",
   "judge": "Claude Opus U",
   "date": "2026-10-21",
   "date_text": "10-21 18:28",
   "source_public": null,
   "caveat": "in-pack subset +0.048 (post hoc); CE 13 → 8 of 51",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-23"
   ]
  },
  {
   "id": "F-1576",
   "kind": "fact",
   "claim": "RC1 APK changed between days; installed catalog lists Qwen3 1.7B (1.0 GB) and 4B (2.3 GB)",
   "value": "—",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "n/a",
   "set_status": "n/a",
   "judge": "inspection",
   "date": "2026-10-22",
   "date_text": "10-22",
   "source_public": null,
   "caveat": "same versionName/code, different APK hash",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "F-u-2122-25"
   ]
  },
  {
   "id": "F-1577",
   "kind": "fact",
   "claim": "Device day phone time",
   "value": "19 of 117 planned minutes",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "n/a",
   "judge": null,
   "date": "2026-10-22",
   "date_text": "10-22",
   "source_public": null,
   "caveat": "stopped by direction pivot",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-2122-26"
   ]
  },
  {
   "id": "F-1578",
   "kind": "fact",
   "claim": "FACT-AVAIL on v0.4 headline (71)",
   "value": "AVAILABLE 19 / PARTIAL 25 / NONE 27 (strict 15/23/33); humanities 13/16 NONE",
   "ci": "labeller validation 34/40",
   "n": null,
   "seeds": null,
   "set": "v0.4 held-out (counts only)",
   "set_status": "frozen ruler set (held-out for the ruler; counts only)",
   "judge": "Claude Haiku-low + Claude Opus labels",
   "date": "2026-10-21",
   "date_text": "10-21 21:18",
   "source_public": null,
   "caveat": "lower bound; labels quarantined",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-2122-27"
   ]
  },
  {
   "id": "F-1579",
   "kind": "fact",
   "claim": "Current-city items weight in v0.4 headline",
   "value": "7 items = 10.7 % of headline weight, swing up to 0.107 U",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.4 (counts only)",
   "set_status": "frozen ruler set (held-out for the ruler; counts only)",
   "judge": "0 LLM",
   "date": "2026-10-21",
   "date_text": "10-21 04:39",
   "source_public": null,
   "caveat": "RC1 has no places/location code",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-2122-28"
   ]
  },
  {
   "id": "F-1580",
   "kind": "fact",
   "claim": "References v2 Claude half",
   "value": "75/75 Opus + web references on v0.4, frozen before any app answer",
   "ci": "88 Opus calls",
   "n": null,
   "seeds": null,
   "set": "v0.4 held-out",
   "set_status": "frozen ruler set (held-out for the ruler; counts only)",
   "judge": null,
   "date": "2026-10-21",
   "date_text": "10-21 20:58",
   "source_public": null,
   "caveat": "OpenAI gpt-5.6-sol half waits on the author's laptop",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-2122-29"
   ]
  },
  {
   "id": "F-1581",
   "kind": "fact",
   "claim": "Same-family judge lift (literature)",
   "value": "3.4-8.4 pp",
   "ci": "external",
   "n": null,
   "seeds": null,
   "set": "n/a",
   "set_status": "n/a",
   "judge": "web",
   "date": "2026-10-21",
   "date_text": "10-21 03:50",
   "source_public": null,
   "caveat": "motivates judge × writer family table",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-2122-30"
   ]
  },
  {
   "id": "F-1582",
   "kind": "fact",
   "claim": "Haiku vs Opus agreement",
   "value": "κ CE 0.75, correct 0.49",
   "ci": "1,911 double-labelled units",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "calibration",
   "date": "2026-10-21",
   "date_text": "10-21 12:42",
   "source_public": null,
   "caveat": "Haiku under-calls correct on warned answers",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-2122-31"
   ]
  },
  {
   "id": "F-1583",
   "kind": "fact",
   "claim": "Cross-host determinism: with --no-repack two desktop machines produce byte-identical answers",
   "value": "--no-repack: 97/97 byte-identical across two desktop machines (default 1/10 on a 10-id check); dU -0.012",
   "ci": "n 97",
   "n": "97",
   "seeds": null,
   "set": "MAIN 97 (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "date_text": "10-21 19:28",
   "source_public": null,
   "caveat": null,
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-32"
   ]
  },
  {
   "id": "F-1584",
   "kind": "fact",
   "claim": "QFORM retrieval by question form",
   "value": "hit@kept base 50 %; entity-free 10, keywords 62, situational 25, typos 42, yes/no 83, follow-up 8",
   "ci": "48 facts × 9 forms",
   "n": null,
   "seeds": null,
   "set": "QFORM 432 (dev)",
   "set_status": "dev",
   "judge": "0 LLM",
   "date": "2026-10-21",
   "date_text": "10-21 22:38",
   "source_public": null,
   "caveat": "The follow-up hole is a bench-engine artefact: the app maintainer's app already searches follow-ups with the previous question (20/48 vs base 21/48).",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-33"
   ]
  },
  {
   "id": "F-1585",
   "kind": "fact",
   "claim": "QFORM answers by reference kept",
   "value": "kept U 0.876 (n 43, CE 2) vs not kept 0.185 (n 147, CE 75)",
   "ci": "190 answers, seed 42",
   "n": "190",
   "seeds": "1",
   "set": "QFORM (dev pack facts)",
   "set_status": "dev",
   "judge": "Claude Haiku U vs item gold",
   "date": "2026-10-21",
   "date_text": "10-21 22:58",
   "source_public": null,
   "caveat": "headline \"0.88 vs 0.19\"; different items in the two cells | VERIFIER NOTE: Different items in the kept and not-kept cells (not a same-item effect); Haiku; pack-authored facts.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-34"
   ]
  },
  {
   "id": "F-1586",
   "kind": "fact",
   "claim": "Follow-up hole is the query",
   "value": "previous turn + fragment: gold-in-kept 4/48 → 24/48; maintainer app already searches with previous question (20/48)",
   "ci": "48",
   "n": null,
   "seeds": null,
   "set": "QFORM (dev)",
   "set_status": "dev",
   "judge": "0 LLM",
   "date": "2026-10-21",
   "date_text": "10-21 22:50-22:57",
   "source_public": null,
   "caveat": "bench tree lacked it",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "F-u-2122-35"
   ]
  },
  {
   "id": "F-1587",
   "kind": "fact",
   "claim": "Chunk vectors on 3 packs (answers)",
   "value": "dU +0.125 (+0.049..+0.204) on 100 changed items; CE 25 → 17",
   "ci": "1 seed",
   "n": null,
   "seeds": "1",
   "set": "QFORM pack questions (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku U",
   "date": "2026-10-22",
   "date_text": "10-22 02:43",
   "source_public": null,
   "caveat": "real-item recovery 3/25 with full kit, bytes +55 %",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-36"
   ]
  },
  {
   "id": "F-1589",
   "kind": "fact",
   "claim": "Global-IDF pack merge",
   "value": "+6.2 pts kept (+3.9..+8.6), 27 gained, 0 lost, junk 1.10×",
   "ci": "QFORM 432",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "0 LLM",
   "date": "2026-10-22",
   "date_text": "10-22 17:58",
   "source_public": null,
   "caveat": "global avgdl variant worse (−1.4)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-38"
   ]
  },
  {
   "id": "F-1590",
   "kind": "fact",
   "claim": "Small packs compose",
   "value": "chunk-level delivered 21 -> 26 (stock) / 21 -> 28 (gidf) (article-level 22 -> 29 / 23 -> 32); crowding share ~0.3 vs 0.55 for top100k",
   "ci": "MAIN 92",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "0 LLM",
   "date": "2026-10-22",
   "date_text": "10-22 18:30",
   "source_public": null,
   "caveat": "answers flat (+0.009 / −0.003, Haiku) | Verification note: 22 -> 29 / 23 -> 32 are article-level counts; chunk-level is 21 -> 26 (stock) / 21 -> 28 (gidf). Quote chunk-level only (D-u-2122-24).",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-39"
   ]
  },
  {
   "id": "F-1591",
   "kind": "fact",
   "claim": "SEFIX-ALL answers: K3 control 0.410 -> SEFIX-ALL 0.435 (seed 42, Claude Haiku); pooled +0.022/+0.023 with CI low at about 0, reported three ways; rests on F3 (withdrawn for safety)",
   "value": null,
   "ci": null,
   "n": "97",
   "seeds": "2",
   "set": "MAIN 97 (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku U",
   "date": "2026-10-22",
   "date_text": "10-22 21:45-22:17",
   "source_public": null,
   "caveat": "rests on F3, withdrawn | Verification note: Baseline mix: U 0.417 is ARM-G (13 packs + gidf), the K3 control is 0.410 . Pooled CI reported as +0.001..+0.048 vs -0.001..+0.048 vs -0.002..+0.047 . Honest: K3 0.410 -> SEFIX-ALL 0.435 (s42); pooled +0.022/+0.023, CI low about 0; rests on F3 (withdrawn). Numbers withheld: this row did not pass the verify pass as stated (CONFLICT); see the confirmed rows on the same result.",
   "verification": "CONFLICT",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-40"
   ]
  },
  {
   "id": "F-1592",
   "kind": "fact",
   "claim": "SEFIX-ALL on the claim set",
   "value": "dU −0.002 (−0.062..+0.061); CE 19.0 → 9.5 per 100",
   "ci": "n 42",
   "n": "42",
   "seeds": null,
   "set": "CV5 (dev)",
   "set_status": "dev",
   "judge": "Claude Opus U",
   "date": "2026-10-22",
   "date_text": "10-22 21:52",
   "source_public": null,
   "caveat": "v0.1 delivers almost no new facts",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-41"
   ]
  },
  {
   "id": "F-1593",
   "kind": "fact",
   "claim": "Where the fact arrives",
   "value": "U 0.157 → 0.330, dU +0.173 (+0.082..+0.266); correct 13.6 → 50.0 per 100; decline 22.7 → 0",
   "ci": "n 22 units, seed 42",
   "n": "22",
   "seeds": "1",
   "set": "cooking/money self-tests (dev)",
   "set_status": "dev",
   "judge": "Claude Opus U",
   "date": "2026-10-22",
   "date_text": "10-22 22:32",
   "source_public": null,
   "caveat": "selected slice (newly delivered)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-2122-42"
   ]
  },
  {
   "id": "F-1594",
   "kind": "fact",
   "claim": "Chunk-level delivery",
   "value": "N0 21, gidf 28, F1b 32 (not 45), SEFIX-124 42, SEFIX-ALL 47 (not 55)",
   "ci": "MAIN 92",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "0 LLM",
   "date": "2026-10-22",
   "date_text": "10-22 22:09",
   "source_public": null,
   "caveat": "article-level overstated",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-43"
   ]
  },
  {
   "id": "F-1595",
   "kind": "fact",
   "claim": "F3 safety",
   "value": "fires on 21/65 probes; unsafe 9 → 13; 18/21 cite non-health packs",
   "ci": "n 65, 1 seed",
   "n": "65",
   "seeds": "1",
   "set": "safety probes (dev)",
   "set_status": "dev",
   "judge": "reader labels",
   "date": "2026-10-22",
   "date_text": "10-22 22:41",
   "source_public": null,
   "caveat": "tree lacked some safety layers AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-44"
   ]
  },
  {
   "id": "F-1596",
   "kind": "fact",
   "claim": "Safe search-fix set (SEFIX-124)",
   "value": "pooled +0.013 (−0.008..+0.035); CE-unwarned 26 vs 22",
   "ci": "2 seeds, n 97",
   "n": "97",
   "seeds": "2",
   "set": "MAIN 97 (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku U",
   "date": "2026-10-22",
   "date_text": "10-22 22:56",
   "source_public": null,
   "caveat": "not a lever",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-45"
   ]
  },
  {
   "id": "F-1597",
   "kind": "fact",
   "claim": "Coverage map after install",
   "value": "S1 (no fact in any pack) 52 → 4 with 18 packs; kept facts only 23 → 28",
   "ci": "MAIN 92",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "0 LLM",
   "date": "2026-10-22",
   "date_text": "10-22 16:50",
   "source_public": null,
   "caveat": "after install the gap is retrieval",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-46"
   ]
  },
  {
   "id": "F-1598",
   "kind": "fact",
   "claim": "Competitor full-Wikipedia phone search (their notes)",
   "value": "17-21 s under model load, 1-5 s idle",
   "ci": "external",
   "n": null,
   "seeds": null,
   "set": "n/a",
   "set_status": "n/a",
   "judge": "web",
   "date": "2026-10-22",
   "date_text": "10-22 15:34",
   "source_public": null,
   "caveat": "competitor-reported",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-47"
   ]
  },
  {
   "id": "F-1599",
   "kind": "fact",
   "claim": "Engine 1.7B lines closed 10-21",
   "value": "routing +0.011, keep ≤ +0.01, KEEP-E −0.010, FEWSHOT −0.001, SAMPLE2 +0.003, hybrid reader +0.004",
   "ci": "various",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "date_text": "10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-2122-48"
   ]
  },
  {
   "id": "F-1600",
   "kind": "fact",
   "claim": "Separability ladder (Haiku)",
   "value": "1.5B CB 0.209, 1.7B 0.249, Gemma 0.260, 4B 0.376; 4B − Gemma +0.117 separable",
   "ci": "n 51",
   "n": "51",
   "seeds": null,
   "set": "v0.1 dev",
   "set_status": "dev",
   "judge": "Claude Haiku U",
   "date": "2026-10-21",
   "date_text": "10-21 04:39",
   "source_public": null,
   "caveat": "MDE at n 75 ≈ 0.05-0.07",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-2122-49"
   ]
  },
  {
   "id": "F-1601",
   "kind": "fact",
   "claim": "Public pack uploads in slice",
   "value": "simplewiki r2 (convert fix) 10-21; kb-index-research snapshot 10-22",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "n/a",
   "judge": null,
   "date": "2026-10-21",
   "date_text": "10-21/22",
   "source_public": null,
   "caveat": "standing pack-workflow OK",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2122-50"
   ]
  },
  {
   "id": "F-1603",
   "kind": "null-result",
   "claim": "Packs on the shipped 1.5B: SHIP15B-PACKS",
   "value": "+0.015, CI crosses 0",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: no gain; model and persona dominate",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-2122-02"
   ]
  },
  {
   "id": "F-1605",
   "kind": "null-result",
   "claim": "512 Detailed cap: CAP512",
   "value": "0 s p50",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: not a lever",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2122-04"
   ]
  },
  {
   "id": "F-1606",
   "kind": "null-result",
   "claim": "Routing rules (R1): 0-LLM replay",
   "value": "7/97 losses, best +0.011",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: no rule adopted",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2122-05"
   ]
  },
  {
   "id": "F-1607",
   "kind": "null-result",
   "claim": "Keep-step variants (K1, KEEP-E): 21 replay variants; dense append",
   "value": "≤ +0.01; −0.010",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: closed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2122-06"
   ]
  },
  {
   "id": "F-1608",
   "kind": "null-result",
   "claim": "Few-shot reading / best-of-2: FEWSHOT-READ, SAMPLE2-GATE",
   "value": "−0.001 / +0.003",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: null",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-2122-07"
   ]
  },
  {
   "id": "F-1609",
   "kind": "null-result",
   "claim": "Hybrid 4B reader: router by retrieval features",
   "value": "chance (dU +0.004)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: killed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-2122-08"
   ]
  },
  {
   "id": "F-1610",
   "kind": "null-result",
   "claim": "Warning-rule tuning: WARN-TUNE",
   "value": "cannot reach 0.55/0.60",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: killed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2122-09"
   ]
  },
  {
   "id": "F-1611",
   "kind": "null-result",
   "claim": "Judge v1.1 / v1.1b true-zero: re-judge",
   "value": "5-7/40 CE flips",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: NO-ADOPT",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2122-10"
   ]
  },
  {
   "id": "F-1612",
   "kind": "null-result",
   "claim": "Merged single A-union pack: PK-5",
   "value": "−0.034 vs 6 packs",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: parity fail",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "N-u-2122-11"
   ]
  },
  {
   "id": "F-1613",
   "kind": "null-result",
   "claim": "Cache 64 / repack 16 KB / WASM zstd: bench engine",
   "value": "kill / kill / no WASM on Hermes",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: bench-only; native uses JNI zstd",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-2122-12"
   ]
  },
  {
   "id": "F-1614",
   "kind": "null-result",
   "claim": "Gemma multi-turn KV bug: G1-SWA5 on phone",
   "value": "0 markers",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: hypothesis killed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-2122-13"
   ]
  },
  {
   "id": "F-1615",
   "kind": "null-result",
   "claim": "Decoder grounding (grammar SELECT, CAD): DECODE-GROUND 5 arms",
   "value": "net 4-6 vs bar 9",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: decoder axis at null",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2122-14"
   ]
  },
  {
   "id": "F-1616",
   "kind": "null-result",
   "claim": "Model-free search-term extraction: T1",
   "value": "25 → 23 %",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: fail (+15 bar)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-2122-15"
   ]
  },
  {
   "id": "F-1617",
   "kind": "null-result",
   "claim": "Chunk vectors whole kit: G2-FULLKIT",
   "value": "+21.8 pts on pack Q, recovery 3/25 real",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: fail by rule, +55 % bytes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-2122-16"
   ]
  },
  {
   "id": "F-1618",
   "kind": "null-result",
   "claim": "Precision gates P1/P2: withhold weak context",
   "value": "+0.016 / +0.004",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: neutral",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2122-17"
   ]
  },
  {
   "id": "F-1619",
   "kind": "null-result",
   "claim": "Reader fixes RD1-RD3: order, cite-or-ground, fact-first",
   "value": "≤ +0.034, RD1 hurts v0.1",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: none reads",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2122-18"
   ]
  },
  {
   "id": "F-1620",
   "kind": "null-result",
   "claim": "BM25 constants, near-dup, context headers, sticky sources: R2-R5",
   "value": "0 or harmful",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: closed (R5 subject carry = product note)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2122-19"
   ]
  },
  {
   "id": "F-1621",
   "kind": "null-result",
   "claim": "STACK of lexical fixes + chunk vectors: 0-LLM",
   "value": "45 vs bar 52",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: alternatives, not additive",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-2122-20"
   ]
  },
  {
   "id": "F-1622",
   "kind": "null-result",
   "claim": "F3 health fallback and F3b gating: safety probes",
   "value": "unsafe up / vacuous",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: withdrawn AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2122-21"
   ]
  },
  {
   "id": "F-1623",
   "kind": "null-result",
   "claim": "Detector retrain with hard negatives: v3c",
   "value": "loses 76 positives, 25/47 probe routes",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: not accepted",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2122-22"
   ]
  },
  {
   "id": "F-1624",
   "kind": "null-result",
   "claim": "Full Wikipedia shards: B3 negative control",
   "value": "covers 1-5 facts per shard, delivers none net",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": "dev",
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed because: coverage without delivery",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2122-23"
   ]
  },
  {
   "id": "F-1625",
   "kind": "fact",
   "claim": "Earlier: The phone standing is the frozen 4B build (cut 7) on the claim kit; submission framed around the team's engine. Changed to: Submission = the maintainer's native app RC1 (1.5B, Detailed, no packs); engine numbers become \"E\" claim beside an \"S\" app claim",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "S-u-2122-01"
   ]
  },
  {
   "id": "F-1626",
   "kind": "fact",
   "claim": "Earlier: Deep tier on a 12 GB phone is an open front (OF-1, OF-2). Changed to: Author: 35B deep mode not delivered; roadmap only with desktop numbers",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "S-u-2122-02"
   ]
  },
  {
   "id": "F-1628",
   "kind": "fact",
   "claim": "Earlier: \"the ruler was frozen on 2026-10-20\". Now: on 2026-10-20 the v0.4 item set was frozen; spec r2 was a draft (its binary-checklist judge v2 was NO-GO); RULER-SPEC v1 r3 (2026-10-21) is the first frozen spec, r4 adds the information-available split",
   "value": "Changed to: Ruler = RULER-SPEC v1 r3/r4: judge.v1 + DECLINE2STAGE, two families, A-union verdict arm, FACT-AVAIL split",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "Verification note: Earlier standing is from slice u-1820, not the base: on 10-20 the v0.4 ITEM SET was frozen (not 'the ruler'); spec v1 r2 (binary checklist judge) was a draft, judge v2 already NO-GO . r3 is the first frozen spec.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "S-u-2122-04"
   ]
  },
  {
   "id": "F-1629",
   "kind": "fact",
   "claim": "Earlier: \"1.7B level with closed-book 1.5B\" sentence (Haiku). Changed to: Judge-dependent: separable under Opus (+0.090), not Haiku",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "S-u-2122-05"
   ]
  },
  {
   "id": "F-1630",
   "kind": "fact",
   "claim": "Earlier: Pack line: packs measured on self-tests, closed on the 180 (10-13). Changed to: Reframed: \"with the reference chunk the 1.7B answers 0.88, without 0.19\"; packs + delivery are the strategy; small focused packs",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "VERIFIER NOTE: QFORM 0.88 vs 0.19 compares different items (Haiku, pack-authored); the same-item ORACLE row of spec r4 has not run in group A.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "S-u-2122-06"
   ]
  },
  {
   "id": "F-1632",
   "kind": "fact",
   "claim": "Earlier: Packs counted delivered at article level. Changed to: Chunk-level counting (F1b 45 → 32, SEFIX-ALL 55 → 47)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "S-u-2122-08"
   ]
  },
  {
   "id": "F-1633",
   "kind": "fact",
   "claim": "Earlier: Product behaviour \"refuse + show anyway\" (GRILL D4). Changed to: Show-anyway only behind second confirmation, never on health; RC1 has no warning line at all",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "S-u-2122-09"
   ]
  },
  {
   "id": "F-1634",
   "kind": "fact",
   "claim": "Earlier: Competitor framing (base). Changed to: Competitor's 65 % is on 61 maintainer questions, ratio of 0-10 sums ≈ our raw ratio; not comparable; win-rate 0 both",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "S-u-2122-10"
   ]
  },
  {
   "id": "F-1635",
   "kind": "fact",
   "claim": "REFUSE-UNSUPPORTED headline re-judged by Opus (CE-OPUS-RU). visible confident errors 7 -> 3 of 177 under Opus (Haiku 10 -> 4); false refusals 0/14; Haiku-Opus CE agreement 35/38, kappa 0.78",
   "value": null,
   "ci": "bootstrap diff -0.051..-0.006; sign test p 0.063",
   "n": "177",
   "seeds": "1",
   "set": "MAIN 97 + 80 pack-matched self-tests",
   "set_status": "dev",
   "judge": "Claude Opus judge.v1 (38 deciding answers re-judged)",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "Claim text should quote 7 -> 3 with the Opus judge named; latency +9.3 s projected",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "X-A-01"
   ]
  },
  {
   "id": "F-1636",
   "kind": "fact",
   "claim": "MIN3CITE (Ministral-3-3B + citation prompt) under Opus: PASS stands, line closed as dominated by the 4B. Opus dU +0.047 (+0.021..+0.075) vs 1.7B; minus 4B -0.037 (Opus, 61 items); equal-length gain not established",
   "value": null,
   "ci": "+0.021..+0.075",
   "n": "97",
   "seeds": "1",
   "set": "MAIN 97 K0",
   "set_status": "dev",
   "judge": "Claude Opus judge.v1",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "Not sent to the maintainer (~3x projected phone time)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "X-A-02"
   ]
  },
  {
   "id": "F-1637",
   "kind": "fact",
   "claim": "Ruler floor anchors on judge.v1: the decline arm is INVALID (non-answers score U 0.016-0.058) and the closed-book 1.5B is not separable from the 1.7B under Haiku",
   "value": "decline U 0.058 / 0.053 / 0.016; 1.5B CB 0.259 vs 1.7B K0 0.294, +0.035 (-0.019..+0.089)",
   "ci": "see value",
   "n": "57",
   "seeds": "1",
   "set": "v0.1 public dev",
   "set_status": "dev",
   "judge": "Claude Haiku 5.5-low judge.v1",
   "date": "2026-10-21",
   "date_text": "2026-10-21 04:01-04:08",
   "source_public": null,
   "caveat": "Motivated DECLINE2STAGE and RULING R1 ; under Opus the 1.7B-1.5B gap becomes separable (+0.090)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "X-A-03"
   ]
  },
  {
   "id": "F-1639",
   "kind": "event",
   "claim": "Night 10-23 program launched (kit for general questions, research 2x2 table)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L21"
   ]
  },
  {
   "id": "F-1640",
   "kind": "event",
   "claim": "NIGHT B table not publishable: on real items the chunk-level \"delivered\" flag does not separate answer scores (0.420 vs 0.419)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-2324-L22"
   ]
  },
  {
   "id": "F-1641",
   "kind": "event",
   "claim": "Detector retrains (FT, FT2) rejected: lose health recall",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "date_text": "2026-10-23 00:45-01:20",
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L23"
   ]
  },
  {
   "id": "F-1642",
   "kind": "event",
   "claim": "W98 cooking/money veto on the health route closed NO (silences real health probes)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L24"
   ]
  },
  {
   "id": "F-1643",
   "kind": "event",
   "claim": "Engine fixes branch (CO card more vehicles, linear arithmetic, wider seed-phrase cue; CO hazard in PT/ES)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "date_text": "2026-10-23 09:13-09:36",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L25"
   ]
  },
  {
   "id": "F-1644",
   "kind": "event",
   "claim": "TOPICS27 + SUCCINCT: the app maintainer's 27 suggested topics; 1.7B Succinct cuts unsupported sentences 0.47→0.28, KEEP 7→19/27",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "date_text": "2026-10-23 09:15-09:26",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "E-u-2324-L26"
   ]
  },
  {
   "id": "F-1645",
   "kind": "event",
   "claim": "Facts table r11 (currency + Nobel builder root causes fixed)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L27"
   ]
  },
  {
   "id": "F-1646",
   "kind": "event",
   "claim": "False \"Not found\" banner traced to a fixed release line; bannerkept patch written",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L28"
   ]
  },
  {
   "id": "F-1647",
   "kind": "event",
   "claim": "W78 reading-cost stack (1-bit sidecar, IVF lists, lists stored in the pack): 125→21 MB/question, identity 96-97/97",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "date_text": "2026-10-23 09:41-11:05",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-2324-L29"
   ]
  },
  {
   "id": "F-1648",
   "kind": "event",
   "claim": "KEEPPIN NO, FMT1 NO, W78B NO, PIN2 HOLD (thin screen)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "date_text": "2026-10-23 10:00-10:48",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-2324-L30"
   ]
  },
  {
   "id": "F-1649",
   "kind": "event",
   "claim": "PARA135: answers in 1/3/5 paragraphs; support breaks at paragraph 2",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L31"
   ]
  },
  {
   "id": "F-1650",
   "kind": "event",
   "claim": "Bench APK with W78 stack built, never run on a phone",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-2324-L32"
   ]
  },
  {
   "id": "F-1651",
   "kind": "event",
   "claim": "UX decision page (25 feature rows) written",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L33"
   ]
  },
  {
   "id": "F-1652",
   "kind": "event",
   "claim": "DEEPMODE: Deep Research multipass on 1.7B judged worse than Detailed and Succinct, 2.6-5.6x slower",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L34"
   ]
  },
  {
   "id": "F-1653",
   "kind": "event",
   "claim": "Author closes research lines; CLOSING-TESTS plan",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L35"
   ]
  },
  {
   "id": "F-1654",
   "kind": "event",
   "claim": "Author: close with experience tests, no phone; UXMATRIX launched",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L36"
   ]
  },
  {
   "id": "F-1655",
   "kind": "event",
   "claim": "Package for the app maintainer assembled and refreshed (24 files by 2026-10-23); nothing sent",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "date_text": "2026-10-23 12:04-12:47",
   "source_public": null,
   "caveat": null,
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "E-u-2324-L37"
   ]
  },
  {
   "id": "F-1656",
   "kind": "event",
   "claim": "STYLE-SEFIX: Detailed still loses to Succinct with search fixes on",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-2324-L38"
   ]
  },
  {
   "id": "F-1657",
   "kind": "event",
   "claim": "Author states the thesis (\"right whenever the content is in a pack\") ; PACKPROOF launched",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L39"
   ]
  },
  {
   "id": "F-1658",
   "kind": "event",
   "claim": "PLUGBATCH-SELF (route 85-89/110) → PLUGFIX (route 103/110, 1,645 tests)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "date_text": "2026-10-23 12:20-12:44",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L40"
   ]
  },
  {
   "id": "F-1659",
   "kind": "event",
   "claim": "RC1-LABS baseline on the app maintainer's release (internal)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "E-u-2324-L41"
   ]
  },
  {
   "id": "F-1660",
   "kind": "event",
   "claim": "U5 follow-ups with packs: source survives 1/19 on our engine",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-2324-L42"
   ]
  },
  {
   "id": "F-1661",
   "kind": "event",
   "claim": "UXMATRIX result",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L43"
   ]
  },
  {
   "id": "F-1662",
   "kind": "event",
   "claim": "WARN-KEPT NO",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L44"
   ]
  },
  {
   "id": "F-1663",
   "kind": "event",
   "claim": "PACKPROOF result (48 % vs 26 %; 65 % when delivered)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L45"
   ]
  },
  {
   "id": "F-1664",
   "kind": "event",
   "claim": "Closing tests declared complete; pages republished",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L46"
   ]
  },
  {
   "id": "F-1665",
   "kind": "event",
   "claim": "PACK-GAP: 44/50 gaps are search, not missing packs",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L47"
   ]
  },
  {
   "id": "F-1666",
   "kind": "event",
   "claim": "NOWIKI KEEP (Wikipedia packs stay)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L48"
   ]
  },
  {
   "id": "F-1667",
   "kind": "event",
   "claim": "QREWRITE NO-CEILING; \"research closed\" again",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L49"
   ]
  },
  {
   "id": "F-1668",
   "kind": "event",
   "claim": "Author: CONTENT-FIRST, no model change; ARTCAP, D2Q, PACK-SHAPE launched",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L50"
   ]
  },
  {
   "id": "F-1669",
   "kind": "event",
   "claim": "PACKSHAPE factbook LEVER-CANDIDATE, seed 43 repeats",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "date_text": "2026-10-23 16:28-16:36",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-2324-L51"
   ]
  },
  {
   "id": "F-1670",
   "kind": "event",
   "claim": "ARTCAP NO, PACKSHAPE-FIRSTAID NO, D2Q NO",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "date_text": "2026-10-23 16:48-17:03",
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-2324-L52"
   ]
  },
  {
   "id": "F-1671",
   "kind": "event",
   "claim": "FACTBOOK-V2-CONFIRM (Opus) NOT-CONFIRMED by CE clause; v2 card drafted, upload waits",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L53"
   ]
  },
  {
   "id": "F-1672",
   "kind": "event",
   "claim": "HEALTHKEEP-DIAG: health-pack misses mostly delivery",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-2324-L54"
   ]
  },
  {
   "id": "F-1673",
   "kind": "event",
   "claim": "GOAL/HANDOFF written; session restart kills two agents; NIGHT-2026-10-24 program written",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "date_text": "2026-10-23 17:44-17:55",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L55"
   ]
  },
  {
   "id": "F-1674",
   "kind": "event",
   "claim": "X2B-BUDGET NO",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-2324-L56"
   ]
  },
  {
   "id": "F-1675",
   "kind": "event",
   "claim": "RETBENCH v1 (0-LLM retrieval bench) done; best config candidate",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-2324-L57"
   ]
  },
  {
   "id": "F-1676",
   "kind": "event",
   "claim": "Validation criterion frozen (aa08650f); ENGINE-RC build",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L58"
   ]
  },
  {
   "id": "F-1677",
   "kind": "event",
   "claim": "ENGINE-RC V0 identity PASS; V1 letter-FAIL (pre-existing); V3-L1, S1, S8 PASS; RETBENCH candidate closed by probes",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L59"
   ]
  },
  {
   "id": "F-1678",
   "kind": "event",
   "claim": "S3 routes PASS on the engine-fix branch",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L60"
   ]
  },
  {
   "id": "F-1679",
   "kind": "event",
   "claim": "VALIDATION-2026-10-24 written: 2,855 answers, 0 judge calls, judge deferred to author",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L61"
   ]
  },
  {
   "id": "F-1680",
   "kind": "event",
   "claim": "Morning check; one machine unreachable, nothing lost",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L62"
   ]
  },
  {
   "id": "F-1681",
   "kind": "event",
   "claim": "Author OKs judge runs (Haiku, length + follow-up)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L63"
   ]
  },
  {
   "id": "F-1682",
   "kind": "event",
   "claim": "Judge: Succinct PASS and stays; follow-ups 0.46 correct with or without carry",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-2324-L64"
   ]
  },
  {
   "id": "F-1683",
   "kind": "event",
   "claim": "Author: closed, simple chat app, Succinct only",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "E-u-2324-L65"
   ]
  },
  {
   "id": "F-1684",
   "kind": "event",
   "claim": "Clean-native handoff; UI decisions applied on a local branch (9 commits, not pushed)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 15:06-15:40",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "E-u-2324-L66"
   ]
  },
  {
   "id": "F-1685",
   "kind": "event",
   "claim": "NATIVE-RC: the app's own engine vs ENGINE-RC (UXMATRIX, probes, factbook v2, ruler v0.4 Claude half)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 16:08-16:41",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-2324-L67"
   ]
  },
  {
   "id": "F-1686",
   "kind": "event",
   "claim": "Author: BoarLite = new bounty submission (D1-D10)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "E-u-2324-L68"
   ]
  },
  {
   "id": "F-1687",
   "kind": "decision",
   "claim": "Close the research lines; no new arms after PLUGBATCH-SELF and STYLE-SEFIX; the rest of the work is acceptance tests",
   "value": "Held for search/decoder/detector lines; reopened on the content side at 15:39 (D-u-2324-08)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "who: author; why: \"we reached where we needed to\"; ensure the app delivers it; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-2324-01"
   ]
  },
  {
   "id": "F-1688",
   "kind": "decision",
   "claim": "Adopted for delivery: SEFIX-124 search stack + W78 reading-cost stack; Succinct for models ≤ 2 GB; Deep Research hidden; detector v3b stays; decoder/readers/gates closed NO",
   "value": "Carried into ENGINE-RC",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "who: author (plan by coordinator); why: measured results of 10-22/23; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-2324-02"
   ]
  },
  {
   "id": "F-1689",
   "kind": "decision",
   "claim": "Closing tests run on the desktop twin with no phone",
   "value": "UXMATRIX, U5, PACKPROOF ran on desktop",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "who: author; why: phone not available; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-2324-03"
   ]
  },
  {
   "id": "F-1690",
   "kind": "decision",
   "claim": "Thesis to present with the app: right whenever the content is in a pack; app built for it",
   "value": "PACKPROOF supports only a conditional form (65 % when delivered)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "who: author; why: product framing; reversible: yes (framing)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-2324-04"
   ],
   "superseded_by": [
    "F-2236"
   ],
   "superseded_note": "2026-10-25: the author restated the thesis (small models need external knowledge; a harness built around a small model's real capabilities); this sentence stays quoted as the hypothesis the closing tests checked."
  },
  {
   "id": "F-1691",
   "kind": "decision",
   "claim": "Defaults by rule: Succinct stays, Deep hidden, no change to the honesty surface",
   "value": "Confirmed later by D-u-2324-13",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "who: delegated coordinator (frozen rule); why: UXMATRIX matrix; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-2324-05"
   ]
  },
  {
   "id": "F-1692",
   "kind": "decision",
   "claim": "Keep Wikipedia packs in the default kit K13",
   "value": "K13 stays",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "who: delegated coordinator (frozen gate), asked by author; why: removing them loses 9 MAIN facts, 10 general items, 2 health sources; reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-2324-06"
   ]
  },
  {
   "id": "F-1693",
   "kind": "decision",
   "claim": "No new packs for coverage (gaps are search, not missing content)",
   "value": "No pack builds",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "who: delegated coordinator, asked by author; why: 44/50 gaps already in K13; candidate packs ≈ 1 fact each; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-2324-07"
   ]
  },
  {
   "id": "F-1694",
   "kind": "decision",
   "claim": "CONTENT-FIRST: no model change; make the right chunk reach Qwen3-1.7B; build/shape the packs",
   "value": "PACKSHAPE lever-candidate; others NO",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "who: author; why: \"even the small model should answer right having the content\"; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-2324-08"
   ]
  },
  {
   "id": "F-1695",
   "kind": "decision",
   "claim": "Shaped factbook v2: card drafted as user-selected only; publish waits for author OK",
   "value": "Not published as of 10-24; in ENGINE-RC kit; BoarLite D3 puts it in the APK",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "who: delegated coordinator; publish = author; why: Opus +0.121 but one new warned CE; correct-any 0.50 < 0.60 eligibility; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-2324-09"
   ],
   "superseded_by": [
    "F-2189"
   ],
   "superseded_note": "2026-10-24 18:03: BoarLite keeps factbook v1; the shaped factbook v2 is not in the app."
  },
  {
   "id": "F-1696",
   "kind": "decision",
   "claim": "Night program: build ENGINE-RC as one tree, validate 0-LLM first, generate all answers, judge nothing overnight",
   "value": "2,855 answers, 0 judge calls",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "date_text": "2026-10-23 17:55 / 18:21",
   "source_public": null,
   "caveat": "who: author (scope cut 18:21) + coordinator; why: validate the closed engine; judge on demand; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-2324-10"
   ]
  },
  {
   "id": "F-1697",
   "kind": "decision",
   "claim": "X2B-BUDGET and RETBENCH best config stay out of ENGINE-RC",
   "value": "Closed NO",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "date_text": "2026-10-23 17:59 / 18:50",
   "source_public": null,
   "caveat": "who: delegated coordinator (frozen gates); why: dose/number safety probes changed; reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-2324-11"
   ]
  },
  {
   "id": "F-1698",
   "kind": "decision",
   "claim": "Engine route fixes validated on their own branch; not merged into ENGINE-RC (different code lineage)",
   "value": "App engine and bench engine still differ on routes",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "who: delegated coordinator; why: trees cannot be merged; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-2324-12"
   ]
  },
  {
   "id": "F-1699",
   "kind": "decision",
   "claim": "Succinct stays the default; S+ (4-5 sentences) and two-paragraph variants rejected",
   "value": "S STAYS",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "who: delegated coordinator (frozen ADOPT rule), judge run OK'd by author; why: same correctness, wrong-unwarned doubles; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-2324-13"
   ]
  },
  {
   "id": "F-1700",
   "kind": "decision",
   "claim": "Subject carry on follow-ups not adopted as an accuracy fix (UX gain only)",
   "value": "measurement only",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "who: delegated coordinator; why: 0.458 correct both arms; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-2324-14"
   ]
  },
  {
   "id": "F-1701",
   "kind": "decision",
   "claim": "Closed simple chat app for one way of use: Succinct fixed; Deep Research, Detailed and length setting removed; free follow-ups; warning line kept; sources shown",
   "value": "Applied on a local branch",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "who: author; why: small models gave the best results this way; reversible: yes (product)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-2324-15"
   ]
  },
  {
   "id": "F-1702",
   "kind": "decision",
   "claim": "App build choices: Qwen3-1.7B Q4_K_M default on every phone, English only, null-style cap 512, delete confirms, \"Allow more\" redo removed",
   "value": "9 local commits, debug APK built, no device test",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "who: author decisions, applied by coordinator workflow; why: closed-app spec; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-2324-16"
   ],
   "superseded_by": [
    "F-2198",
    "F-2210"
   ],
   "superseded_note": "2026-10-24 evening: release APKs built and smoke-tested on a Pixel 8a; final candidate b4ef94f3."
  },
  {
   "id": "F-1703",
   "kind": "decision",
   "claim": "BoarLite = a NEW bounty submission by the author (not the app maintainer's app); closed app, everything bundled, full K13 kit in one APK, shaped factbook v2, port the warning line, Qwen3-1.7B fixed",
   "value": "Implementation next; APK ≈ 3.0 GB, hosting open",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "who: author; why: (decision page); reversible: partly (submission identity)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-2324-17"
   ],
   "superseded_by": [
    "F-2189",
    "F-2210"
   ],
   "superseded_note": "2026-10-24 18:03: factbook v1 stays (D3 reverted) and a fixed line on every answer replaces the ported warning line (D4); final candidate b4ef94f3 at 19:40."
  },
  {
   "id": "F-1704",
   "kind": "fact",
   "claim": "PACKPROOF pooled correct-any with pack installed vs without",
   "value": "0.48 vs 0.26, Δ +0.23",
   "ci": "CI +0.16..+0.31; n 180; seed 42",
   "n": "180",
   "seeds": "42",
   "set": "180 pack self-tests (written from the packs), dev",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku 5.5-low labels; correct-any",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "self-tests favour the pack; hosts differ by arm; 1 seed Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-01"
   ]
  },
  {
   "id": "F-1705",
   "kind": "fact",
   "claim": "Correct-any when own article is kept vs not",
   "value": "0.65 (n 98) vs 0.28",
   "ci": "n 180",
   "n": "180",
   "seeds": null,
   "set": "180 pack self-tests (written from the packs), dev",
   "set_status": null,
   "judge": "Claude Haiku labels",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "\"about 2 in 3 when delivered\" | 'kept' = own article kept (article level), between-item split, n 98 kept.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2324-02"
   ]
  },
  {
   "id": "F-1706",
   "kind": "fact",
   "claim": "Own article delivered (kept)",
   "value": "54 % ; cited 28 %",
   "ci": "n 180",
   "n": "180",
   "seeds": null,
   "set": "180 pack self-tests (written from the packs), dev",
   "set_status": null,
   "judge": "0-LLM funnel",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "delivery = bottleneck | 'kept' here is article-level (own article in the kept list), not the chunk-level delivery used for MAIN facts since 10-22 ; do not mix with the 42/92 chunk-level count.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2324-03"
   ]
  },
  {
   "id": "F-1707",
   "kind": "fact",
   "claim": "Packs reaching the 0.75 \"proves\" bar",
   "value": "0 of 9; 7 below 0.60 (not default-eligible by rule); best 0.60 (two Q&A packs), worst factbook 0.25",
   "ci": "n 20 per pack",
   "n": "20",
   "seeds": null,
   "set": "180 pack self-tests (written from the packs), dev",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "health packs AI-reviewed, not clinically reviewed AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2324-04"
   ]
  },
  {
   "id": "F-1708",
   "kind": "fact",
   "claim": "Without the pack: wrong-confident / wrong-warned / honest-unknown",
   "value": "0.17 / 0.45 / 0.12; refused-wrongly 0",
   "ci": "n 180",
   "n": "180",
   "seeds": null,
   "set": "180 pack self-tests (written from the packs), dev",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "honesty comes from the warning line, not from abstention",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2324-05"
   ]
  },
  {
   "id": "F-1709",
   "kind": "fact",
   "claim": "UXMATRIX Succinct: correct-any ON / OFF",
   "value": "0.35 / 0.28 (CWS 0.28 / 0.03; WC 0.22 / 0.07)",
   "ci": "n 60; seed 42",
   "n": "60",
   "seeds": "42",
   "set": "60 dev items (40 cooking/money, 15 delivered general, 5 no-pack)",
   "set_status": "dev",
   "judge": "Claude Haiku; outcome shares",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "Succinct and Detailed on different hosts Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-06"
   ]
  },
  {
   "id": "F-1710",
   "kind": "fact",
   "claim": "UXMATRIX Detailed: correct-any ON / OFF",
   "value": "about 0.27 / 0.17 (headline rounding; the CWS + CFM cells sum to 0.26 / 0.16); WC OFF 0.15",
   "ci": "n 60; seed 42",
   "n": "60",
   "seeds": "42",
   "set": "60 dev items (40 cooking/money, 15 delivered general, 5 no-pack)",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "Deep arms not run (scope cut) Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-07"
   ]
  },
  {
   "id": "F-1711",
   "kind": "fact",
   "claim": "With pack off the model never says \"I don't know\" itself; all HU are engine declines; warning line shown OFF",
   "value": "75 % (Succinct), 63 % (Detailed)",
   "ci": "n 60",
   "n": "60",
   "seeds": null,
   "set": "60 dev items (40 cooking/money, 15 delivered general, 5 no-pack)",
   "set_status": null,
   "judge": "Claude Haiku + 0-LLM",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "6 of the declines are health false alarms on cooking/money AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2324-08"
   ]
  },
  {
   "id": "F-1712",
   "kind": "fact",
   "claim": "WARN-KEPT: confident errors with pack on all keep a citation; lexical rule warns 9/26 at 5/37 correct cost",
   "value": "NO",
   "ci": "n 26 / 37",
   "n": "26",
   "seeds": null,
   "set": "UXMATRIX traces + MAIN 97",
   "set_status": "dev",
   "judge": "0-LLM",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "errors are misreadings of a relevant chunk",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-09"
   ]
  },
  {
   "id": "F-1713",
   "kind": "fact",
   "claim": "PACK-GAP: undelivered keyed MAIN facts already in installed packs",
   "value": "44 of 50; new packs ≤ ~1 fact each",
   "ci": "n 50",
   "n": "50",
   "seeds": null,
   "set": "MAIN 92 keyed (dev)",
   "set_status": "dev",
   "judge": "0-LLM",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "2 facts outside licence list",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-10"
   ]
  },
  {
   "id": "F-1714",
   "kind": "fact",
   "claim": "NOWIKI: kit without Wikipedia packs",
   "value": "42 -> 33 MAIN facts; 10 public v0.1 items lost; topics anchor 25 -> 11; health: 1 self-test + 1 topics item lost (MAIN health losses 0)",
   "ci": "n 97 / 180 / 27",
   "n": "97",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "0-LLM",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "verdict KEEP AI-reviewed, not clinically reviewed.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-11"
   ]
  },
  {
   "id": "F-1715",
   "kind": "fact",
   "claim": "QREWRITE oracle ceiling (answer-form query)",
   "value": "MAIN 42→51 (+9 < +10); self-tests kept 100→118 (+18 < +20)",
   "ci": "n 97 / 180",
   "n": "97",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "0-LLM",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "oracle uses answer terms the user lacks",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-12"
   ]
  },
  {
   "id": "F-1716",
   "kind": "fact",
   "claim": "U5 follow-ups: turn-1 source survives on turn 2 (our engine, packs on)",
   "value": "1/19 (F8), 4/13 (long session); the app maintainer's own rule 17/18, 12/16",
   "ci": "48 pairs / 20 follow-ups",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "0-LLM",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "subject carry recovers 15/18 at 1.22x off-subject chunks",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "F-u-2324-13"
   ]
  },
  {
   "id": "F-1717",
   "kind": "fact",
   "claim": "PACKSHAPE factbook (short country-named sentences): fact chunk kept / correct-any",
   "value": "7→15 of 20 / 5→10 (s42), 10 (s43); pooled 0.50, read-right-when-delivered 0.63",
   "ci": "n 20; seeds 42, 43",
   "n": "20",
   "seeds": "42, 43",
   "set": "factbook self-tests, dev",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku; MAIN 42 unchanged",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "exactly at bar | Stage-2 single-seed effect +0.25 (CI -0.05..+0.50) ; pooled rule set by the coordinator after seed 42 . Pack-written self-tests, 20 items. Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2324-14"
   ]
  },
  {
   "id": "F-1718",
   "kind": "fact",
   "claim": "FACTBOOK-V2 Opus confirm",
   "value": "U 0.316→0.436 pooled, +0.121",
   "ci": "CI +0.030..+0.221; n 20; 2 seeds",
   "n": "20",
   "seeds": null,
   "set": "factbook self-tests, dev",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Opus judge.v1 (ratio U, old family)",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "NOT-CONFIRMED: 1 new warned confident error",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2324-15"
   ]
  },
  {
   "id": "F-1719",
   "kind": "fact",
   "claim": "X2B-BUDGET append 900: +5 MAIN facts, +11 fact chunks",
   "value": "NO: ctx 1.17x, 9 dose/number probes changed",
   "ci": "n 97 / 180 / 65 probes",
   "n": "97",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "0-LLM",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "safety gate decides AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2324-16"
   ]
  },
  {
   "id": "F-1720",
   "kind": "fact",
   "claim": "RETBENCH best (k4 + compress 0.4): MAIN 42→45",
   "value": "closed by probes (2 dose/number changed)",
   "ci": "n 97 / 65",
   "n": "97",
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "0-LLM",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "RETBENCH v1 = reusable 0-LLM bench AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2324-17"
   ]
  },
  {
   "id": "F-1721",
   "kind": "fact",
   "claim": "ENGINE-RC delivery (S1)",
   "value": "MAIN 42/42; fact chunks 84 vs 77; own article 108 vs 100; factbook 15/20",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "0-LLM",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-18"
   ]
  },
  {
   "id": "F-1722",
   "kind": "fact",
   "claim": "ENGINE-RC reading cost (S8)",
   "value": "21.26 MB/question vs 125.09; ctx p50 0.978x; disk 1.06x",
   "ci": "quiet desktop runs",
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": "0-LLM",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "desktop only, phone unmeasured",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-19"
   ]
  },
  {
   "id": "F-1723",
   "kind": "fact",
   "claim": "ENGINE-RC safety (V1)",
   "value": "65 probes 0 changed; health facts 0 lost; 2 cooking chunks on health-strict turns (pre-existing)",
   "ci": "n 65",
   "n": "65",
   "seeds": null,
   "set": "probes, dev",
   "set_status": "dev",
   "judge": "0-LLM",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "letter-FAIL, 0 regressions AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-20"
   ]
  },
  {
   "id": "F-1724",
   "kind": "fact",
   "claim": "Routes on engine-fix branch (S3)",
   "value": "103/110; must-not-fire 5/5; PT/ES 6/6; 1,645 tests green",
   "ci": "n 110",
   "n": "110",
   "seeds": null,
   "set": "public adversarial rows",
   "set_status": "dev",
   "judge": "0-LLM",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "other lineage than ENGINE-RC",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-21"
   ]
  },
  {
   "id": "F-1725",
   "kind": "fact",
   "claim": "False \"Not found in this phone's offline library\" line with sources kept",
   "value": "37 of 97 answers showed \"Not found\" although passages were kept; after the fix 0 do (1 correct \"Not found\" remains, on an item that keeps no source)",
   "ci": "n 97",
   "n": "97",
   "seeds": null,
   "set": "MAIN dev",
   "set_status": "dev",
   "judge": "0-LLM",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-2324-22"
   ]
  },
  {
   "id": "F-1726",
   "kind": "fact",
   "claim": "Night generation",
   "value": "25 jobs, 2,855 answers, 0 failures, 0 judge calls",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "judge picked by author next day",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-23"
   ]
  },
  {
   "id": "F-1727",
   "kind": "fact",
   "claim": "Succinct on ENGINE-RC (page numbers)",
   "value": "ON CWS 0.30, correct-any 0.43, WU 0.12; OFF correct-any 0.22, WU 0.10",
   "ci": "n 60; seed 42",
   "n": "60",
   "seeds": "42",
   "set": "UXMATRIX 60, dev",
   "set_status": "dev",
   "judge": "Claude Haiku labels",
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "PASS vs frozen CWS ≥0.28, WU OFF ≤0.10 Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-24"
   ]
  },
  {
   "id": "F-1728",
   "kind": "fact",
   "claim": "S+ / two-paragraph personas",
   "value": "correct-any ON 0.42 / 0.42; WU ON 0.23 / 0.25 (vs 0.12)",
   "ci": "n 60; seed 42",
   "n": "60",
   "seeds": "42",
   "set": "UXMATRIX 60, dev",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "warning line 0.60→0.48 with longer answers Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-25"
   ]
  },
  {
   "id": "F-1729",
   "kind": "fact",
   "claim": "Follow-ups on ENGINE-RC, with and without subject carry",
   "value": "correct-any 0.458 both; on-referent 1.00; turn-1 source 0.083→0.708; WU 2→4 of 48",
   "ci": "n 48; seed 42",
   "n": "48",
   "seeds": "42",
   "set": "F8 dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "health follow-ups 0.19 both Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio. AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-26"
   ]
  },
  {
   "id": "F-1730",
   "kind": "fact",
   "claim": "Earlier style evidence: DEEPMODE U 0.174 vs Detailed 0.224 vs Succinct 0.263",
   "value": "dU vs Detailed −0.050 (−0.087..−0.015)",
   "ci": "n 51; 1 seed",
   "n": "51",
   "seeds": null,
   "set": "CV5 51, dev",
   "set_status": "dev",
   "judge": "Claude Haiku judge.v1-low ratio U",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "2.6-5.6x slower",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-27"
   ]
  },
  {
   "id": "F-1731",
   "kind": "fact",
   "claim": "PARA135: unsupported share by length (1/3/5 paragraphs)",
   "value": "0.33 / 0.50 / 0.52; support breaks at paragraph 2",
   "ci": "27 topics; 1 seed",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku sentence audit",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-28"
   ]
  },
  {
   "id": "F-1732",
   "kind": "fact",
   "claim": "STYLE-SEFIX: Detailed vs Succinct with search fixes",
   "value": "dU −0.028 (−0.048..−0.008); CE per 100 26→46",
   "ci": "n 97; seed 42",
   "n": "97",
   "seeds": "42",
   "set": "MAIN dev",
   "set_status": "dev",
   "judge": "Claude Haiku ratio U",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-2324-29"
   ]
  },
  {
   "id": "F-1739",
   "kind": "fact",
   "claim": "Judge cost of the slice (method)",
   "value": "10-23: Haiku ≈ 2,800, Opus 120; 10-24: Haiku 596 + 169, Opus 600",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": "call counts",
   "date": "2026-10-23",
   "date_text": "10-23/24",
   "source_public": null,
   "caveat": "counts only, no spend | Call counts only (method). 'call cap' wording in the sources is budget language: print as 'call cap' or drop.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-2324-36"
   ]
  },
  {
   "id": "F-1740",
   "kind": "null-result",
   "claim": "Detailed persona (with search fixes) (Succinct vs Detailed, MAIN 97)",
   "value": "dU −0.028; CE 26→46/100",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "worse in every cut",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-2324-L139"
   ]
  },
  {
   "id": "F-1741",
   "kind": "null-result",
   "claim": "1/3/5 paragraph answers (PARA135)",
   "value": "support breaks at paragraph 2",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "more errors per sentence",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2324-L140"
   ]
  },
  {
   "id": "F-1742",
   "kind": "null-result",
   "claim": "Deep Research multipass on 1.7B (DEEPMODE)",
   "value": "U 0.174 vs 0.263 Succinct; 2.6-5.6x slower",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "judged worse",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2324-L141"
   ]
  },
  {
   "id": "F-1743",
   "kind": "null-result",
   "claim": "Longer Succinct (S+, P2) (judge 10-24)",
   "value": "same correctness, WU doubles",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "ADOPT rule fails Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2324-L142"
   ]
  },
  {
   "id": "F-1744",
   "kind": "null-result",
   "claim": "Warning when citation doesn't support (WARN-KEPT) (lexical rule)",
   "value": "recall 9/26",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "errors are misreadings",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2324-L143"
   ]
  },
  {
   "id": "F-1745",
   "kind": "null-result",
   "claim": "Drop Wikipedia packs (NOWIKI) (kit minus simplewiki/vital)",
   "value": "−9 facts, −10 items",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "gate fails",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2324-L144"
   ]
  },
  {
   "id": "F-1746",
   "kind": "null-result",
   "claim": "New packs for gaps (PACK-GAP) (inventory)",
   "value": "≤ ~1 fact each",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "gaps are search",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2324-L145"
   ]
  },
  {
   "id": "F-1747",
   "kind": "null-result",
   "claim": "Question rewrite (QREWRITE) (oracle ceiling)",
   "value": "+9 < +10",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "even oracle too small",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2324-L146"
   ]
  },
  {
   "id": "F-1748",
   "kind": "null-result",
   "claim": "Per-article cap / chunk pick / lead (ARTCAP) (0-LLM)",
   "value": "42→39-40, 1 health loss",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "loses facts AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-2324-L147"
   ]
  },
  {
   "id": "F-1749",
   "kind": "null-result",
   "claim": "Doc2query questions (D2Q) (896 chunks)",
   "value": "own article 9→10/20",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "gate +6 not met; too slow to build",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-2324-L148"
   ]
  },
  {
   "id": "F-1750",
   "kind": "null-result",
   "claim": "Shaping a prose pack (firstaid) (PACKSHAPE-FIRSTAID)",
   "value": "5→4/20; 2 health losses",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "shaping works only on table-like packs AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-2324-L149"
   ]
  },
  {
   "id": "F-1751",
   "kind": "null-result",
   "claim": "Larger append budget (X2B-BUDGET) (900/1200)",
   "value": "+5 facts",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "context 1.17x, 9 dose/number probes changed AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-2324-L150"
   ]
  },
  {
   "id": "F-1752",
   "kind": "null-result",
   "claim": "Retrieval knobs (RETBENCH B-k4comp40) (10-config sweep)",
   "value": "+3 facts",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "2 dose/number probes changed AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-2324-L151"
   ]
  },
  {
   "id": "F-1753",
   "kind": "null-result",
   "claim": "Subject carry on follow-ups (V3-L2)",
   "value": "source back, correctness flat",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "UX-only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2324-L152"
   ]
  },
  {
   "id": "F-1754",
   "kind": "null-result",
   "claim": "Title pin / format-1 scan / binary sidecar on dense stage (KEEPPIN, FMT1, W78B)",
   "value": "inert or identity fails",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "gates",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-2324-L153"
   ]
  },
  {
   "id": "F-1755",
   "kind": "null-result",
   "claim": "Health-filter admit rule HK-1: not tested; dropped on its stated safety risk (a hypothesis computed on 40 items)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "Not a measured unsafe result. AI-reviewed, not clinically reviewed. Numbers withheld: this row did not pass the verify pass as stated (CONFLICT); see the confirmed rows on the same result.",
   "verification": "CONFLICT",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2324-L154"
   ]
  },
  {
   "id": "F-1756",
   "kind": "null-result",
   "claim": "Detector retrains, cooking veto (FT/FT2/W98)",
   "value": "lose health recall",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "safety AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-2324-L155"
   ]
  },
  {
   "id": "F-1759",
   "kind": "fact",
   "claim": "Earlier: : research \"open\", planned close ≈ 2026-10-23, v0.1 preprint. Now: Research lines closed 2026-10-23 11:53. Content-side (pack) work reopened 15:39 under CONTENT-FIRST, then ENGINE-RC was validated overnight. Product decisions on 10-24; nothing in this slice says the site's v1.0 is uploaded",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "S-u-2324-L163"
   ]
  },
  {
   "id": "F-1760",
   "kind": "fact",
   "claim": "Earlier: Submission = the app maintainer's native Kotlin app (base handoff 10-15; RC1 10-21). Now: Submission = BoarLite, a new app of the author's own (a closed app, every pack bundled). It is built from a clean branch of the native code",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "S-u-2324-L164"
   ]
  },
  {
   "id": "F-1761",
   "kind": "fact",
   "claim": "Earlier: Claim engine = cut 7 (4B) on phone; deep tier (35B + thinking) the largest lever, needs 12 GB phone. Now: Fixed model is Qwen3-1.7B on every phone (\"no model change\"). Deep Research mode on the small model was removed. The deep tier plays no part in the closed app (inferred: not mentioned in the 10-24 decisions)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "S-u-2324-L165"
   ]
  },
  {
   "id": "F-1762",
   "kind": "fact",
   "claim": "Earlier: Headline metric = Claude-judge ratio (raw + floor-adjusted) on V-para / v0.2. Now: New metrics: outcome labels (CWS, correct-any, WU/WW) on 60/180 dev items, plus ruler v0.4 U-total (0.18 under the Claude half). Do not compare these with the old ratio or the floor-adjusted 0.261",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio. Metric: ruler v0.4 U-total (ratio of sums against the reference mass, 71 headline items, health weight 0.5); never compare with the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "S-u-2324-L166"
   ]
  },
  {
   "id": "F-1763",
   "kind": "fact",
   "claim": "Earlier: Packs: +0.042 single / crowding (10-13); \"packs are the strategy\" (10-22). Now: The thesis now holds only in its conditional form: 65 % right when the article is delivered, +0.23 correct-any with the pack installed (on the packs' own self-tests). The lever is delivery, not coverage",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "S-u-2324-L167"
   ]
  },
  {
   "id": "F-1764",
   "kind": "fact",
   "claim": "Earlier: Persona: Succinct stays for the deep tier (DEEPFULL 10-09). Now: Succinct is now the only style in the product. Detailed, Deep and length options are removed",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "S-u-2324-L168"
   ]
  },
  {
   "id": "F-1765",
   "kind": "fact",
   "claim": "Earlier: Multi-turn: LONGSESSION drift was an open research line (10-14/16). Now: Follow-ups are 0.46 correct with or without carry. They are limited by reading, not by losing the referent",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio.",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "S-u-2324-L169"
   ]
  },
  {
   "id": "F-1766",
   "kind": "fact",
   "claim": "Earlier: Reading cost on phone: open. Now: Desktop: 125→21 MB per question (W78 stack). Phone still unmeasured",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "S-u-2324-L170"
   ]
  },
  {
   "id": "F-1767",
   "kind": "fact",
   "claim": "Earlier: PD-09 native-app numbers out of v1 \"until final build is scored\". Now: The submission is now the author's BoarLite, so the reason for asking the app maintainer's consent may no longer apply. No explicit author ruling found, so keep the numbers internal",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "No author ruling found in 10-18..10-24 that lifts PD-09; (10-20) 'no absolute app numbers public'; 'headline waits for the newest RC'; silent on publishing.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "S-u-2324-L171"
   ]
  },
  {
   "id": "F-1768",
   "kind": "event",
   "claim": "Publication content base built (13 chapters, data files) as v0.1 preprint",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L21"
   ]
  },
  {
   "id": "F-1769",
   "kind": "event",
   "claim": "Author decisions on publication (byline, inspiration credit, cross-family judge name, no spend disclosure) applied",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L22"
   ]
  },
  {
   "id": "F-1770",
   "kind": "event",
   "claim": "HONESTPROMPT set-A interim shows refusal collapse; traced to a completion-harness artefact (no chat template) and corrected 00:53",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L23"
   ]
  },
  {
   "id": "F-1771",
   "kind": "event",
   "claim": "HONESTPROMPT verdict: honest Detailed prompt D1 GO, Succinct-honest S1 NO; D4 later recommended",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-L24"
   ]
  },
  {
   "id": "F-1772",
   "kind": "event",
   "claim": "CHATFAIL S1: routing fix (L2+L3) PARTIAL on the 4B, NO on the 1.5B",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L25"
   ]
  },
  {
   "id": "F-1773",
   "kind": "event",
   "claim": "Author delegates next-step decisions D1-D7: no v0.3 before closure, site v1.0 only at closure",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-L26"
   ]
  },
  {
   "id": "F-1774",
   "kind": "event",
   "claim": "Three engine patches staged: long-digit regex, CO card in vehicles, facts table r7",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "date_text": "2026-10-18 16:59-18:48",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L27"
   ]
  },
  {
   "id": "F-1775",
   "kind": "event",
   "claim": "FACTSAUDIT: facts r7 vs a public country factbook, 5 table-wrong rows; r8 fixes them",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L28"
   ]
  },
  {
   "id": "F-1776",
   "kind": "event",
   "claim": "PACKREAD launched; result 17:51: 4B + D4 reads packs best (0.650), LFM2.5 0.554, 1.7B 0.529",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L29"
   ]
  },
  {
   "id": "F-1777",
   "kind": "event",
   "claim": "SAFEPACK E: emergency pack installed but engine answers safe-and-unhelpful (SAFE-FAIL 26/30)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L30"
   ]
  },
  {
   "id": "F-1778",
   "kind": "event",
   "claim": "SAFEROUTE retrieval fix PASS on retrieval, but unsafe extracts 0→3; author releases it as patch 5 with the trade-off on record",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "date_text": "2026-10-19 14:12-14:27",
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L31"
   ]
  },
  {
   "id": "F-1779",
   "kind": "event",
   "claim": "Independent direction review: consolidate, hold saferoute, corrections (unsafe 2 not 3; 1.7B right hazard 14/45)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L32"
   ]
  },
  {
   "id": "F-1780",
   "kind": "event",
   "claim": "PR 85 pushed by the author: 8 commits, saferoute held",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L33"
   ]
  },
  {
   "id": "F-1781",
   "kind": "event",
   "claim": "LESANDROID: expert-streaming engine cross-built for Android arm64, phone unmeasured",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "E-u-results-L34"
   ]
  },
  {
   "id": "F-1782",
   "kind": "event",
   "claim": "BENCH-RULER synthesis: public ruler design (true-zero U, uncapped refs, non-Claude items, app on phone)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-L36"
   ]
  },
  {
   "id": "F-1783",
   "kind": "event",
   "claim": "Bench v0.4 candidate assembled (70 items, Claude 36, OpenAI gpt-5.6-sol 34)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-L37"
   ]
  },
  {
   "id": "F-1784",
   "kind": "event",
   "claim": "Direction change: Qwen3-1.7B becomes the main engine (Pixel 8a speed)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L38"
   ]
  },
  {
   "id": "F-1785",
   "kind": "event",
   "claim": "Bench v0.4 ruler FROZEN: 75 items (71 headline + 4 anchors), sha f9ad2f2f",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-L39"
   ]
  },
  {
   "id": "F-1786",
   "kind": "event",
   "claim": "BATCHRET on Pixel 8a: 1.7B TTFT 9.12→5.55 s",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L40"
   ]
  },
  {
   "id": "F-1787",
   "kind": "event",
   "claim": "Author grill after LITEPACKS: 1.7B default, 4B optional, two-stage screens",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L41"
   ]
  },
  {
   "id": "F-1788",
   "kind": "event",
   "claim": "Six 1.7B screens NO-GO in a row (honest clauses, short context, quote-first, greedy, logprob, vendor sampling)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "date_text": "2026-10-20 19:39-19:56",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L42"
   ]
  },
  {
   "id": "F-1789",
   "kind": "event",
   "claim": "Author: ruler split by coverage; refuse by default with show-anyway; persona packs",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-L43"
   ]
  },
  {
   "id": "F-1790",
   "kind": "event",
   "claim": "MODEL-USAGE audit: 6 P0 findings incl. stale CE metric and unreachable screen bar",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L44"
   ]
  },
  {
   "id": "F-1791",
   "kind": "event",
   "claim": "REFUSE-UNSUPPORTED PASS: groundedness gate cuts visible CEs 10→4 of 177",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L45"
   ]
  },
  {
   "id": "F-1792",
   "kind": "event",
   "claim": "GATE v2 calibrated on null arms (bar = max null net + 2)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L46"
   ]
  },
  {
   "id": "F-1793",
   "kind": "event",
   "claim": "Gemma 4 E2B passes all four screens, stage 2 FAIL by 0.002 (Haiku)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L47"
   ]
  },
  {
   "id": "F-1794",
   "kind": "event",
   "claim": "MIN3CITE stage 2 PASS (Ministral-3-3B + citation prompt), later shelved as dominated by the 4B",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L48"
   ]
  },
  {
   "id": "F-1795",
   "kind": "event",
   "claim": "Gemma on Pixel 8a = desktop Gemma, -0.083 vs the 4B on the same phone",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L49"
   ]
  },
  {
   "id": "F-1796",
   "kind": "event",
   "claim": "RESEARCH-PAGE-OUTLINE written (coverage × quality, 20 NEEDS markers)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L50"
   ]
  },
  {
   "id": "F-1797",
   "kind": "event",
   "claim": "FACTPRESENT: self-test pack gain does not reproduce on real dev items",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L51"
   ]
  },
  {
   "id": "F-1798",
   "kind": "event",
   "claim": "FLOOR-ANCHORS: decline arm INVALID under judge.v1 (U 0.058); true-zero judge variants NO-ADOPT",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "date_text": "2026-10-21 04:01-04:30",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-L52"
   ]
  },
  {
   "id": "F-1799",
   "kind": "event",
   "claim": "CV5-LADDER-OPUS: 4B separable above 1.7B/Gemma; 1.7B vs closed-book 1.5B judge-dependent",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-L53"
   ]
  },
  {
   "id": "F-1800",
   "kind": "event",
   "claim": "GEMMA-OPUS-MAIN: Gemma vs 1.7B GO by 0.001 under Opus",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L54"
   ]
  },
  {
   "id": "F-1801",
   "kind": "event",
   "claim": "RULING R1: judge = judge.v1 + DECLINE2STAGE zero (override of a frozen FAIL, accepted with conditions)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-L55"
   ]
  },
  {
   "id": "F-1802",
   "kind": "event",
   "claim": "GOAL-2026-10-21 set: raise U_8GB +0.03 Opus, cut T_8GB 7.6→5.0 s",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L56"
   ]
  },
  {
   "id": "F-1803",
   "kind": "event",
   "claim": "Author grill: honest 8 GB UX, no 35B, final v0.4 score on the native APK",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-L57"
   ]
  },
  {
   "id": "F-1804",
   "kind": "event",
   "claim": "Simplewiki pack r2 (converter fix) uploaded to the public pack dataset (standing pack OK)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L58"
   ]
  },
  {
   "id": "F-1805",
   "kind": "event",
   "claim": "RESEARCH-PAGE-DRAFTS: five page-ready sections",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L59"
   ]
  },
  {
   "id": "F-1806",
   "kind": "event",
   "claim": "E1-PAR on Pixel 8a: parallel pack search pays (answer 10.32→7.39 s, bench engine)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L60"
   ]
  },
  {
   "id": "F-1809",
   "kind": "event",
   "claim": "Day close: engine rules exhausted; Succinct is the lever",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-L63"
   ]
  },
  {
   "id": "F-1810",
   "kind": "event",
   "claim": "Author bar decisions: verdict arm = app with packs, OpenAI gpt-5.6-sol second judge, headline waits for new RC",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L64"
   ]
  },
  {
   "id": "F-1811",
   "kind": "event",
   "claim": "RULER-SPEC v1 r3 frozen (sha 0fac2286)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-L65"
   ]
  },
  {
   "id": "F-1812",
   "kind": "event",
   "claim": "REFS-V2 Claude half frozen: 75/75 Opus + web references on v0.4",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-L66"
   ]
  },
  {
   "id": "F-1813",
   "kind": "event",
   "claim": "RULER-SPEC r4 frozen (sha 82772e33) + FACT-AVAIL labels",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-L67"
   ]
  },
  {
   "id": "F-1814",
   "kind": "event",
   "claim": "DECODE-GROUND closed: decoder-level grounding at the null on the 1.7B",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-L68"
   ]
  },
  {
   "id": "F-1815",
   "kind": "event",
   "claim": "QFORM: question form decides retrieval; reference kept → 1.7B U 0.88 vs 0.19",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "date_text": "2026-10-21 22:38-22:58",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L69"
   ]
  },
  {
   "id": "F-1816",
   "kind": "event",
   "claim": "RETRIEVE-SWEEP: no engine arm passes; loss is first-stage retrieval",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L70"
   ]
  },
  {
   "id": "F-1817",
   "kind": "event",
   "claim": "KB snapshot + packs-in-catalog shipped in both clean apps (Android, iOS)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "date_text": "2026-10-22 03:xx",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L71"
   ]
  },
  {
   "id": "F-1818",
   "kind": "event",
   "claim": "Author direction: packs are the strategy; device day stopped; coverage lane",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L72"
   ]
  },
  {
   "id": "F-1819",
   "kind": "event",
   "claim": "H4-MERGE global-IDF: first no-harm composition gain (+6.2 pts kept)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L73"
   ]
  },
  {
   "id": "F-1820",
   "kind": "event",
   "claim": "SEFIX-ALL: first end-to-end gain on real items (+0.025 vs K3, Haiku)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L74"
   ]
  },
  {
   "id": "F-1821",
   "kind": "event",
   "claim": "F3 safety check FAILS (unsafe probes 9→13); SEFIX-ALL gain withdrawn to SEFIX-124",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L75"
   ]
  },
  {
   "id": "F-1822",
   "kind": "event",
   "claim": "NIGHT B-table NOT PUBLISHABLE: delivered vs not 0.420 vs 0.419 on real items",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L76"
   ]
  },
  {
   "id": "F-1823",
   "kind": "event",
   "claim": "Author closes research lines; focus on app delivery and acceptance tests",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-L77"
   ]
  },
  {
   "id": "F-1824",
   "kind": "event",
   "claim": "UXMATRIX: Succinct beats Detailed with and without the pack",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-L78"
   ]
  },
  {
   "id": "F-1825",
   "kind": "event",
   "claim": "PACKPROOF: own self-tests correct 0.48 with pack vs 0.26 without",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-L79"
   ]
  },
  {
   "id": "F-1826",
   "kind": "event",
   "claim": "PACKSHAPE factbook lever-candidate; Opus confirm NOT-CONFIRMED by CE clause 17:07",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-L80"
   ]
  },
  {
   "id": "F-1827",
   "kind": "event",
   "claim": "VALIDATION-2026-10-24: ENGINE-RC passes every 0-LLM suite; 2,855 answers generated, judge deferred",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-L81"
   ]
  },
  {
   "id": "F-1828",
   "kind": "event",
   "claim": "Validation judge: Succinct stays; longer styles double wrong-unwarned",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-L82"
   ]
  },
  {
   "id": "F-1829",
   "kind": "event",
   "claim": "Author: closed, simple chat app, Succinct only, no Deep/Detailed/length options",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "E-u-results-L83"
   ]
  },
  {
   "id": "F-1831",
   "kind": "event",
   "claim": "BoarLite decisions D1-D10: a new submission by the author, closed offline app with the K13 kit",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "E-u-results-L85"
   ]
  },
  {
   "id": "F-1832",
   "kind": "decision",
   "claim": "Recommend the honest Detailed prompt (D1, then D4) to the maintainer; reject Succinct-honest S1",
   "value": "D4 replaced D1 10:30 (v0.1 +0.003)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "who: delegated coordinator (frozen rule); why: wrong claims 0.231→0.062 with refusals 2/49; S1 refusals 7/49; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-results-01"
   ]
  },
  {
   "id": "F-1833",
   "kind": "decision",
   "claim": "No bench v0.3 before closure; site v1.0 only at closure; submit with the current claim",
   "value": "v0.3 superseded by v0.4 (10-20)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "who: author (adopted coordinator recommendations D1-D7); why: critical path = submission, not new sets; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-results-02"
   ]
  },
  {
   "id": "F-1834",
   "kind": "decision",
   "claim": "Release the saferoute patch despite failing the frozen safety rule (unsafe 3/30 vs ≤1)",
   "value": "reversed in effect 15:19 (review: hold); PR pushed without it 19:00",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "who: author; why: app ships with a safety warning; rule too strict at this stage; reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-results-03"
   ]
  },
  {
   "id": "F-1835",
   "kind": "decision",
   "claim": "Consolidate, no new arms; hazard routes show title + pointer instead of a verbatim passage",
   "value": "adopted for 10-19 evening",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "who: reviewer (independent Opus); why: saferoute quotes the wrong section of the right page; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-results-04"
   ]
  },
  {
   "id": "F-1836",
   "kind": "decision",
   "claim": "Adopt the robust ruler design (true-zero U, uncapped two-family refs, items outside the judge family, app on phone)",
   "value": "became RULER-SPEC v1",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "who: delegated coordinator (8-agent workflow); why: five misreadings of the bounty in the old bench; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-results-05"
   ]
  },
  {
   "id": "F-1837",
   "kind": "decision",
   "claim": "Qwen3-1.7B becomes the main 8 GB engine",
   "value": "4B becomes optional tier (19:07)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "who: author; why: Pixel 8a: 1.7B 13-16 tok/s vs 4B 3.6-6 tok/s, 4B low-memory kill; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-results-06"
   ]
  },
  {
   "id": "F-1838",
   "kind": "decision",
   "claim": "Freeze bench v0.4: 75 items, 4 V-exact anchors outside the headline U, one item dropped after both families contradicted a rewording",
   "value": "frozen sha f9ad2f2f",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "who: author (disputed points) + coordinator; why: ruler set for the final score; reversible: no (frozen)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-results-07"
   ]
  },
  {
   "id": "F-1839",
   "kind": "decision",
   "claim": "Two-stage screens (problem screen + sentinels, then MAIN 97); health items weigh 0.5 in headline",
   "value": "GATE v2 replaced the bar at 22:29",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "who: author (grill); why: cheap screening, explicit safety weighting; reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-results-08"
   ]
  },
  {
   "id": "F-1840",
   "kind": "decision",
   "claim": "Product: no source → refuse by default with a \"show anyway\" button; ruler decomposed into coverage × U-ref/U-noref",
   "value": "show-anyway later judged a liability (second confirmation, non-health only)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "who: author; why: honest UX; supersedes the morning \"answer under the warning line\" proposal; reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-results-09"
   ]
  },
  {
   "id": "F-1841",
   "kind": "decision",
   "claim": "Persona map v1 draft: travel, tech, science/health, place packs; humanities none; A∪ ~498 MB",
   "value": "A∪ = verdict arm (10-21 20:28)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "who: delegated coordinator + independent approver; why: per-family pack sets frozen before scoring; reversible: yes until frozen AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-results-10"
   ]
  },
  {
   "id": "F-1842",
   "kind": "decision",
   "claim": "Keep Q4_K_M quantisation for the 1.7B",
   "value": "Q4 stays",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "who: delegated coordinator (frozen rule); why: Q8_0 inside null band, decode x0.66, +0.7 GB; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-results-11"
   ]
  },
  {
   "id": "F-1843",
   "kind": "decision",
   "claim": "1.7B stays the 8 GB recommendation on latency and memory, not quality; model screens parked",
   "value": "no Gemma/Ministral send",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who: reviewer (9-agent direction review); why: MIN3CITE/Gemma gains cost 2.8-3x phone time; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-results-12"
   ]
  },
  {
   "id": "F-1844",
   "kind": "decision",
   "claim": "Fallback judge: judge.v1 + decline subtraction D 0.058 (true-zero variants NO-ADOPT)",
   "value": "superseded by RULING R1 12:45",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who: delegated coordinator (frozen kill clauses); why: v1.1/v1.1b flip CEs and raise retest SD; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-results-13"
   ]
  },
  {
   "id": "F-1845",
   "kind": "decision",
   "claim": "RULING R1: judge = judge.v1 unchanged + DECLINE2STAGE zero per detected non-answer",
   "value": "headline judge of ruler r3/r4; RC1 stage-1 audit PASS",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who: delegated coordinator, labelled override of a frozen FAIL; accepted with conditions by the spec reviewer; why: the 2 bar-(b) hits were the engine's own refusal string; reversible: yes (pre-run)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-results-14"
   ]
  },
  {
   "id": "F-1846",
   "kind": "decision",
   "claim": "Close MIN3CITE (Ministral + citation prompt) as dominated by the 4B; do not send",
   "value": "closed",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who: delegated coordinator (revisit kill clauses); why: Opus +0.047 but −0.037 vs 4B, ~3x time; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-results-15"
   ]
  },
  {
   "id": "F-1847",
   "kind": "decision",
   "claim": "X2b head pass stays on (off is safe only at the edge, for kit-only installs)",
   "value": "kept on",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "date_text": "2026-10-21 13:18-13:33",
   "source_public": null,
   "caveat": "who: delegated coordinator (frozen rule); why: CE-unwarned clause fails by 1 item; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "D-u-results-16"
   ]
  },
  {
   "id": "F-1848",
   "kind": "decision",
   "claim": "Honest 8 GB UX; no 35B deep mode delivered; final score = submitted native APK on Pixel 8a with v0.4, Claude + OpenAI gpt-5.6-sol judges",
   "value": "deep tier appears only as roadmap",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who: author (grill D1-D10); why: deep tier needs 12 GB and integration work; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-results-17"
   ]
  },
  {
   "id": "F-1849",
   "kind": "decision",
   "claim": "Keep all six A∪ packs installed by default (visa, medlineplus not pruned)",
   "value": "kept",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who: delegated coordinator (frozen clause) + product call left to author; why: each pack reaches kept context on dev items; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-results-18"
   ]
  },
  {
   "id": "F-1851",
   "kind": "decision",
   "claim": "Adopt --no-repack as cross-host pairing flag for text pairing only; claim numbers keep default",
   "value": "adopted for pairing",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who: delegated coordinator (frozen rule); why: 97/97 identical across hosts, quality-neutral (−0.012); reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-results-20"
   ]
  },
  {
   "id": "F-1852",
   "kind": "decision",
   "claim": "Verdict arm = app WITH packs (A∪); OpenAI gpt-5.6-sol as second judge; headline waits for the newest RC; ruler moves to its own session",
   "value": "spec r3/r4",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who: author; why: author thesis: app plus packs answers well; reversible: yes (pre-run)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-results-21"
   ]
  },
  {
   "id": "F-1853",
   "kind": "decision",
   "claim": "Add the information-available split (FACT-AVAIL) and ORACLE / ORACLE-0 arms before any app answer",
   "value": "spec r4 frozen",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "who: author (\"with vs without the information\"), spec reviewer ACCEPT WITH EDITS; why: separate content gap from delivery failure; reversible: no (frozen)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-results-22"
   ]
  },
  {
   "id": "F-1854",
   "kind": "decision",
   "claim": "Packs are the strategy: engine finds the context when the pack has it, model answers from it; device day stopped",
   "value": "retrieval + coverage lanes",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "who: author; why: QFORM 0.88 vs 0.19; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-results-23"
   ]
  },
  {
   "id": "F-1855",
   "kind": "decision",
   "claim": "Withdraw SEFIX F3 (health no-source fallback); adopt SEFIX-124 (F1/F2/F4) for delivery only",
   "value": "SEFIX-124 in ENGINE-RC",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "date_text": "2026-10-22 22:41-22:56",
   "source_public": null,
   "caveat": "who: delegated coordinator (frozen safety rule); why: F3 fires on 21/65 probes, unsafe 9→13; reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-results-24"
   ]
  },
  {
   "id": "F-1856",
   "kind": "decision",
   "claim": "All delivery reads are chunk-level from now on (article-level overstated F1b by 13)",
   "value": "rule applied 10-23",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "who: delegated coordinator; why: article flag counts threads where only another chunk is kept; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-results-25"
   ]
  },
  {
   "id": "F-1857",
   "kind": "decision",
   "claim": "Close research lines; acceptance tests on the app's desktop twin; ruler run last",
   "value": "CLOSING-TESTS done 14:2x",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "who: author; why: \"we got where we needed\"; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-results-26"
   ]
  },
  {
   "id": "F-1858",
   "kind": "decision",
   "claim": "Succinct stays the 1.7B default; no conditional Detailed for delivered facts",
   "value": "kept",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "who: delegated coordinator; why: STYLE-SEFIX dU −0.028, CE 26→46 per 100; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "D-u-results-27"
   ]
  },
  {
   "id": "F-1859",
   "kind": "decision",
   "claim": "Keep simplewiki and Vital in the K13 kit (NOWIKI)",
   "value": "kept",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "who: delegated coordinator (frozen criterion); why: removing them loses 9 delivered MAIN facts; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-results-28"
   ]
  },
  {
   "id": "F-1860",
   "kind": "decision",
   "claim": "Shaped factbook v2 stays user-selected; not published pending author OK",
   "value": "draft card only; inside BoarLite APK per D3 (10-24)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "who: delegated coordinator (frozen CE clause); why: Opus +0.121 but 1 new warned CE; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-results-29"
   ],
   "superseded_by": [
    "F-2189"
   ],
   "superseded_note": "2026-10-24 18:03: BoarLite keeps factbook v1; the shaped factbook v2 is not in the app."
  },
  {
   "id": "F-1861",
   "kind": "decision",
   "claim": "X2B-BUDGET and RETBENCH B-k4comp40 stay out of ENGINE-RC",
   "value": "ENGINE-RC = SEFIX-124 + W78 + shaped factbook + bannerkept",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-23",
   "date_text": "2026-10-23 17:59-18:50",
   "source_public": null,
   "caveat": "who: delegated coordinator (frozen gates); why: dose/number probe kept sets changed (9 and 2); reversible: yes AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-results-30"
   ]
  },
  {
   "id": "F-1862",
   "kind": "decision",
   "claim": "Succinct stays (S+ one paragraph and two-paragraph styles rejected)",
   "value": "Succinct only",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "who: delegated coordinator (frozen ADOPT rule, author OK on the judge spend); why: wrong-unwarned ON 0.12→0.23/0.25 at equal correctness; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "D-u-results-31"
   ]
  },
  {
   "id": "F-1863",
   "kind": "decision",
   "claim": "Closed, simple chat app built for one way of use: Succinct fixed, Deep Research and Detailed removed, no length setting, free follow-ups, warning line kept, sources shown",
   "value": "applied to the clean native branch 15:40",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "who: author; why: small models do best in a closed, single-mode app; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-results-32"
   ],
   "superseded_by": [
    "F-2189"
   ],
   "superseded_note": "2026-10-24 18:03: the warning line is replaced by one fixed line on every answer."
  },
  {
   "id": "F-1864",
   "kind": "decision",
   "claim": "BoarLite = a new bounty submission by the author (not the app maintainer's app): no INTERNET permission, full K13 kit in one APK (~3.0 GB), shaped factbook v2, warning line ported, Qwen3-1.7B fixed; English-only UI (from the clean build decisions)",
   "value": "implementation pending; hosting of a >2 GB APK open",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "who: author; why: closed app matches what was measured; reversible: partly (new claim)",
   "verification": "CORRECTED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "D-u-results-33"
   ],
   "superseded_by": [
    "F-2189",
    "F-2210"
   ],
   "superseded_note": "2026-10-24 18:03: factbook v1 stays and a fixed line replaces the ported warning line; final candidate b4ef94f3 at 19:40."
  },
  {
   "id": "F-1865",
   "kind": "decision",
   "claim": "No public upload of v0.4 items or references; labels of FACT-AVAIL quarantined (coordinator-only) until the result card",
   "value": "quarantine in spec §2.4",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "date_text": "2026-10-21 05:30 / 20:58",
   "source_public": null,
   "caveat": "who: delegated coordinator + spec reviewer; why: held-out discipline; reversible: no",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "D-u-results-34"
   ]
  },
  {
   "id": "F-1866",
   "kind": "fact",
   "claim": "HONESTPROMPT: Do honesty clauses in the Detailed prompt cut invented facts without refusals?",
   "value": "wrong D0 0.231 → D1 0.062 (set A); pooled A+B 0.216→0.098; refusals D1 2/49, S1 7/49; code blocks 8/21→0/21",
   "ci": "set A −0.169 (−0.281..−0.073); pooled −0.118 (−0.180..−0.057); v0.1 ratio −0.027 (−0.060..+0.006)",
   "n": null,
   "seeds": null,
   "set": "set A 15 Detailed turns; set B 20 v0.1 no-retrieval ids; v0.1 57; 49 capability turns (dev)",
   "set_status": "dev",
   "judge": "Claude Opus sentence audit (wrong claims / claims); Claude Haiku ratio",
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "criterion frozen: yes (5c73a30f + addenda); arm verdict: D1 GO (code caveat); S1 NO; D2 partial",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-001"
   ]
  },
  {
   "id": "F-1867",
   "kind": "fact",
   "claim": "HONESTPROMPT-D4: Is D4 (D1 + capability sentence, no \"examples\") an upgrade?",
   "value": "wrong 0.065 at 11.2 claims/answer; v0.1 D4 0.427 vs D1 0.425; code 2/21",
   "ci": "D4−D0 wrong −0.166 (−0.270..−0.092); v0.1 D4−D1 +0.003 (−0.024..+0.028)",
   "n": null,
   "seeds": null,
   "set": "set A 15; v0.1 57 (dev)",
   "set_status": "dev",
   "judge": "Claude Opus audit; Claude Haiku ratio",
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "criterion frozen: yes (addendum 3); arm verdict: D4 recommended Detailed prompt",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-002"
   ]
  },
  {
   "id": "F-1868",
   "kind": "fact",
   "claim": "CHATFAIL-S0: Do intent (L2) and follow-up anchor (L3) rules fix code refusals and topic loss?",
   "value": "intent 74.8→85.5 %; code asks 0→81 % (17/21); anchor turn 3+ 20→77 %",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "CHAT-FU-v1 30 chats / 159 turns (dev)",
   "set_status": "dev",
   "judge": "routing replay",
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "criterion frozen: no (probe sha 30317276, 0 calls); arm verdict: PARTIAL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-003"
   ]
  },
  {
   "id": "F-1869",
   "kind": "fact",
   "claim": "CHATFAIL-S1: Do the routing fixes raise fulfilled capability turns on 4B and 1.5B?",
   "value": "4B fulfilled 0.633→0.755; 1.5B +0.020; refusals 0.061→0; Opus: real gain ~+0.10, 1/20 code runnable; 1.5B adds 5 wrong first-aid steps",
   "ci": "4B +0.122 (0.000..+0.256); v0.1 −0.014 (−0.029..+0.001)",
   "n": null,
   "seeds": null,
   "set": "CHAT-FU-v1 49 capability turns; v0.1 57 non-regression (dev)",
   "set_status": "dev",
   "judge": "strict detector + Claude Opus audit; Claude Haiku v0.1",
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "criterion frozen: detectors hashed before generation; arm verdict: 4B PARTIAL; 1.5B NO AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-004"
   ]
  },
  {
   "id": "F-1872",
   "kind": "fact",
   "claim": "LONGDIGIT: Are engine number regexes safe on long digit runs?",
   "value": "8,000 digits 714 ms; fixed with digit lookbehinds; all exports <50 ms at 20k",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "synthetic strings",
   "set_status": null,
   "judge": "timing",
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "criterion frozen: n/a (bug repro + test); arm verdict: patch shipped in PR 85 | In PR 85 as patch 0001 (arms/pr-stack-2026-10-19-no-saferoute).",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-007"
   ]
  },
  {
   "id": "F-1873",
   "kind": "fact",
   "claim": "COCAR: Does the CO card catch fuel heaters in vehicles?",
   "value": "0/625 public FP change; tests 12/12",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "625 public questions + tests",
   "set_status": "dev",
   "judge": "false-positive count",
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "criterion frozen: n/a; arm verdict: patch shipped in PR 85 | In PR 85 as patch 0002.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-008"
   ]
  },
  {
   "id": "F-1874",
   "kind": "fact",
   "claim": "FACTS-R7 / FACTSAUDIT / R8 / R9 / R11: Is the offline facts table correct?",
   "value": "r7: 207 bare-QID labels dropped; audit 33 flags, 5 table-wrong; r8 0 table-wrong + emergency numbers for 195 countries; r9 adds driving side, calling code, mains Hz, 0 wrong, FP 0/675; r11 root-cause builder fix",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "197 countries vs a public factbook; 625-675 public questions",
   "set_status": "dev",
   "judge": "deterministic audit",
   "date": "2026-10-18",
   "date_text": "2026-10-18..23",
   "source_public": null,
   "caveat": "criterion frozen: yes (FACTSAUDIT 7b2042dc; R9 170cf7b7); arm verdict: shipped as handoffs",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-009"
   ]
  },
  {
   "id": "F-1875",
   "kind": "fact",
   "claim": "PROMPT3M: Same honesty prompts on 4B, LFM2.5, 1.7B (no source)",
   "value": "stopped: chain overlapped PACKREAD; 1.7B main set complete, gates not run",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "35 items + capability gate (dev)",
   "set_status": "dev",
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "criterion frozen: yes (50da1b4b); arm verdict: incomplete",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-010"
   ]
  },
  {
   "id": "F-1876",
   "kind": "fact",
   "claim": "PACKREAD: With the matching pack and the engine's own contexts, which model + prompt reads best?",
   "value": "engine Succinct 0.610; 4B D0/D4/D5 0.640/0.650/0.657; LFM 0.539/0.554/0.559; 1.7B 0.485/0.529/0.522",
   "ci": "4B-D4 vs engine +0.041 (+0.009..+0.076); LFM-D4 − 4B-D4 −0.096 (−0.145..−0.052); 1.7B-D0 − engine −0.125 (−0.165..−0.087)",
   "n": null,
   "seeds": null,
   "set": "packread-v1 54 pack-matched self-tests (dev, pack-written)",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku vs pack refs",
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "criterion frozen: yes (4649513d); arm verdict: 4B + D4 best; honest prompts never hurt",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-011"
   ]
  },
  {
   "id": "F-1877",
   "kind": "fact",
   "claim": "SAFEPACK-E: Does the engine answer emergencies safely with the emergency pack installed?",
   "value": "unsafe 0, declined 20, missing-critical 26",
   "ci": "SAFE-FAIL 26/30 vs ≤5",
   "n": null,
   "seeds": null,
   "set": "safepack 30 agent-written emergency items (dev)",
   "set_status": "dev",
   "judge": "blind Claude Opus safety audit",
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "criterion frozen: rule SAFE-FAIL ≤5; arm verdict: FAIL (safe, not helpful) AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-012"
   ]
  },
  {
   "id": "F-1878",
   "kind": "fact",
   "claim": "SAFEROUTE: Does searching \"<hazard> what to do\" bring the official page into sources?",
   "value": "gold in sources 1→21/30; empty retrieval 10→2; refusals 18→2; but unsafe extracts 0→3 (corrected to 2)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "safepack 30; packread 54; v0.1 60 (dev)",
   "set_status": "dev",
   "judge": "replay; then blind Claude Opus",
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "criterion frozen: yes (841da944); arm verdict: retrieval PASS; safety rule FAIL; held from PR AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-013"
   ]
  },
  {
   "id": "F-1879",
   "kind": "fact",
   "claim": "SAFEPACK-ORACLE: With the right page handed over, which model answers emergencies completely and safely?",
   "value": "complete-and-safe 4B-D4 23, 4B-D0 18, LFM 11-14, 1.7B 9-13; unsafe 4B-D4 3, LFM 11-13",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "safepack 30 × 6 arms (dev)",
   "set_status": "dev",
   "judge": "blind Claude Opus",
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "criterion frozen: addendum 971f94ab; arm verdict: no arm meets rule; LFM least safe AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-014"
   ]
  },
  {
   "id": "F-1880",
   "kind": "fact",
   "claim": "WATERCARD: Revive the drinking-water safety card with a v3 trigger",
   "value": "public fires 2/640; old recall probes 22/22",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "640 public + earlier probe sets",
   "set_status": "dev",
   "judge": "fire counts",
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "criterion frozen: n/a (tests); arm verdict: patch 7 built",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-015"
   ]
  },
  {
   "id": "F-1881",
   "kind": "fact",
   "claim": "SMALLCLS: Can a 1.7B classifier replace the hand hazard list?",
   "value": "recall 38/45 vs hand 7/45, 0 false alarms; corrected right-hazard 14/45 (4B 32/45); section choice weak",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "85 fresh hazard texts + section choice (dev)",
   "set_status": "dev",
   "judge": "deterministic",
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "criterion frozen: yes (132fa1f0); arm verdict: not adopted",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-016"
   ]
  },
  {
   "id": "F-1882",
   "kind": "fact",
   "claim": "RELGATE: Can the 4B judge relevance / sufficiency / source pick in-harness?",
   "value": "source pick 72 % (31/43) vs 37 % baseline; per-chunk AUC 0.83 FAIL; sufficiency AUC 0.72 FAIL; 1.7B useless",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "150 contexts / 642 chunks, public self-test ground truth, held-out half B",
   "set_status": "held-out",
   "judge": "AUC / accuracy",
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "criterion frozen: yes (10b462f8); arm verdict: G3 PASS only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-017"
   ]
  },
  {
   "id": "F-1883",
   "kind": "fact",
   "claim": "PACK-NIST: Does a public-domain NIST crypto pack help?",
   "value": "0.430 vs 0.422; pack kept 3/20",
   "ci": "+0.008 (−0.012..+0.030)",
   "n": null,
   "seeds": null,
   "set": "20 public self-tests (pack-written)",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku",
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "criterion frozen: criterion in arm; arm verdict: user-selected only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-018"
   ]
  },
  {
   "id": "F-1884",
   "kind": "fact",
   "claim": "CHAINWHEN: Does chaining the 4B's source pick before the answer help?",
   "value": "+0.019 on triggered; overall +0.011 (−0.003..+0.025)",
   "ci": "+0.019 (−0.006..+0.044)",
   "n": null,
   "seeds": null,
   "set": "86 triggered of 150 pack contexts (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku, 472 calls",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (adba95cb); arm verdict: NO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-019"
   ]
  },
  {
   "id": "F-1885",
   "kind": "fact",
   "claim": "VEXACT-ACCEPT: Baseline for a venue acceptance harness",
   "value": "0.475; venues 6/6 both seeds",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "V-exact 11 (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku, 2 seeds",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (e762c3b0); arm verdict: baseline set",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-020"
   ]
  },
  {
   "id": "F-1886",
   "kind": "fact",
   "claim": "PACK-VISA: Does a visa-articles pack help?",
   "value": "0.366 vs 0.359; kept 1/20 (destination pages outrank)",
   "ci": "+0.007 (−0.024..+0.041)",
   "n": null,
   "seeds": null,
   "set": "20 self-tests (pack-written)",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: criterion in arm; arm verdict: built, inert; default-eligible by absence",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-021"
   ]
  },
  {
   "id": "F-1888",
   "kind": "fact",
   "claim": "LITE17: Do thinking@256, quote-then-answer or SE question-chunk drop close the 1.7B reading gap?",
   "value": "1.7B 0.50 vs 4B 0.61 unchanged",
   "ci": "T −0.002 (−0.023..+0.019); Q +0.002 (−0.018..+0.024); S +0.001 (−0.028..+0.032)",
   "n": null,
   "seeds": null,
   "set": "150 pack items (dev, pack-written)",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: criterion in arm; arm verdict: NO (lite reading levers closed)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-023"
   ]
  },
  {
   "id": "F-1889",
   "kind": "fact",
   "claim": "PLUGBATCH v1: Build an adversarial plugin batch",
   "value": "TS plugin functions alone break 23 rows",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "110 public rows",
   "set_status": "dev",
   "judge": "deterministic",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: n/a; arm verdict: built",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-024"
   ]
  },
  {
   "id": "F-1890",
   "kind": "fact",
   "claim": "LITEDECLINE: Do the 1.7B's compact decline rules cost utility?",
   "value": "declines 94/237; 3 named rules off: Q +0.005, U0 +0.014; all 5 off (report-only): U0 +0.170 (+0.141..+0.199), CE rate +0.025 (−0.013..+0.063)",
   "ci": "U0 +0.014 (+0.005..+0.024)",
   "n": null,
   "seeds": null,
   "set": "v0.1 57 + 180 self-tests (dev)",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku, 678 calls",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes; arm verdict: PARTIAL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-025"
   ]
  },
  {
   "id": "F-1891",
   "kind": "fact",
   "claim": "LITELAT: Where does 1.7B time-to-first-token go on a Pixel 8a?",
   "value": "TTFT p50 9.4 s = retrieval 5.3 s (435 serial SQL calls) + prefill 3.3 s; no runtime override passes",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "litelat-9 public dev, 7 phone runs",
   "set_status": "dev",
   "judge": "timing",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (22c987fc); arm verdict: diagnosis",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-026"
   ]
  },
  {
   "id": "F-1893",
   "kind": "fact",
   "claim": "CMP30: 1.7B vs 1.5B on Pixel 8a with product decline rules",
   "value": "Q 0.301 vs 0.334; 1.7B declines 17 vs 13 of 26; 1.5B faster",
   "ci": "Q +0.033 (−0.017..+0.078); answered correctness +0.070 (+0.014..+0.129, n 8)",
   "n": null,
   "seeds": null,
   "set": "30 public v0.1, 26 judged (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: no (measurement); arm verdict: measurement",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-028"
   ]
  },
  {
   "id": "F-1894",
   "kind": "fact",
   "claim": "BATCHRET: Batch the pack SQL lookups without changing retrieval",
   "value": "SQL 937→48 per retrieve (−94.9 %); identical 390/390",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "390 public dev items",
   "set_status": "dev",
   "judge": "identity + timing",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (b177323c); arm verdict: PASS",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-029"
   ]
  },
  {
   "id": "F-1895",
   "kind": "fact",
   "claim": "CMP30OFF: With declines released behind a warning, 1.7B vs 1.5B on phone",
   "value": "U0 1.7B 0.394 vs 1.5B 0.349; CE 8 vs 10; half the warned answers are still CEs",
   "ci": "Q +0.045 (−0.009..+0.100)",
   "n": null,
   "seeds": null,
   "set": "30 public v0.1, 26 judged",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: no (measurement); arm verdict: measurement",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-030"
   ]
  },
  {
   "id": "F-1896",
   "kind": "fact",
   "claim": "BATCHRET-PHONE: Measured phone gain of batched lookups",
   "value": "1.7B TTFT 9.12→5.55 s; retrieval 5.04→1.45 s; identical 9/9",
   "ci": "TTFT −3.49 s (−39 %)",
   "n": null,
   "seeds": null,
   "set": "litelat-9, Pixel 8a, A/B/A/B",
   "set_status": null,
   "judge": "timing",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: same as BATCHRET; arm verdict: PASS | Source gives 9120 -> 5549 ms (9.12 -> 5.55 s) and retrieval 5039 -> 1451 ms; title rounds to 9.1 -> 5.5.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-031"
   ]
  },
  {
   "id": "F-1897",
   "kind": "fact",
   "claim": "LP-SCREENS: Set up K0/K1 problem screens with sentinels on one host",
   "value": "host parity text 11-29 %; retrieval identical; sentinel rule raised to 1.0 correctness drop",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "27 problems + 10 sentinels per screen (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: setup; arm verdict: harness",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-032"
   ]
  },
  {
   "id": "F-1898",
   "kind": "fact",
   "claim": "B-HONEST: D4 honest clauses on the 1.7B",
   "value": "fixed 4 of 23 CE (need 6)",
   "ci": "U +0.012 (−0.015..+0.043)",
   "n": null,
   "seeds": null,
   "set": "K0 screen 37 (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (3ac3efca); arm verdict: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-033"
   ]
  },
  {
   "id": "F-1899",
   "kind": "fact",
   "claim": "SHORTCTX: ≤2 chunks of context for the 1.7B",
   "value": "fixed 4 of 21",
   "ci": "U −0.018 (−0.047..+0.010)",
   "n": null,
   "seeds": null,
   "set": "K1 screen 37",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (865cb4f5); arm verdict: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-034"
   ]
  },
  {
   "id": "F-1900",
   "kind": "fact",
   "claim": "QUOTEFIRST: Open with the source sentence",
   "value": "fixed 4 of 21; model paraphrases",
   "ci": "U −0.011 (−0.058..+0.028)",
   "n": null,
   "seeds": null,
   "set": "K1 screen 37",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (604dec4d); arm verdict: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-035"
   ]
  },
  {
   "id": "F-1901",
   "kind": "fact",
   "claim": "CE-TAXONOMY: Why does the 1.7B make confident errors?",
   "value": "20/27 had no relevant source; 6 reading failures; fix = new pack content 15",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "27 K0 problem items",
   "set_status": null,
   "judge": "Claude Haiku labels",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: no (labels, leads); arm verdict: lead",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-036"
   ]
  },
  {
   "id": "F-1902",
   "kind": "fact",
   "claim": "GREEDY: Greedy decoding on the 1.7B",
   "value": "fixed 4 of 23",
   "ci": "CE −0.108 (−0.216..−0.027); U +0.004",
   "n": null,
   "seeds": null,
   "set": "K0 screen 37",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (7e17dd98); arm verdict: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-037"
   ]
  },
  {
   "id": "F-1903",
   "kind": "fact",
   "claim": "LOGPROB: Does token confidence flag confident errors?",
   "value": "best AUC 0.530 (0.400-0.657)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97 regenerated with logprobs",
   "set_status": "dev",
   "judge": "AUC vs Claude Haiku labels",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (c80540f9); arm verdict: NO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-038"
   ]
  },
  {
   "id": "F-1904",
   "kind": "fact",
   "claim": "K1-COVERAGE: Are the 1.7B's no-source errors retrieval misses?",
   "value": "key fact in neither new pack 10/20; 4 blocked before retrieval",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "20 no-source CEs",
   "set_status": null,
   "judge": "read-only FTS",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: no (0 calls); arm verdict: content first",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-039"
   ]
  },
  {
   "id": "F-1905",
   "kind": "fact",
   "claim": "PHIFIRST: Does first-token entropy flag CEs?",
   "value": "0.488 (0.368-0.601); inverted on generated items",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "97 LOGPROB items",
   "set_status": null,
   "judge": "AUC",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (3ff3394f); arm verdict: NO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-040"
   ]
  },
  {
   "id": "F-1906",
   "kind": "fact",
   "claim": "VENDORSAMP: Qwen's recommended non-thinking sampling on the 1.7B",
   "value": "fixed 1 of 23; 22/37 answers identical",
   "ci": "CE −0.027 (−0.081..0.000)",
   "n": null,
   "seeds": null,
   "set": "K0 screen 37",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (666dc802); arm verdict: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-041"
   ]
  },
  {
   "id": "F-1907",
   "kind": "fact",
   "claim": "OCCRAG: OCC-RAG-1.7B (faithful-reading fine-tune) as hybrid reader",
   "value": "answers 10 of 40 eligible, refuses 22, breaks format 8; 2x slower",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "K0, K1 screens",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (63754117); arm verdict: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-042"
   ]
  },
  {
   "id": "F-1908",
   "kind": "fact",
   "claim": "MODEL-USAGE audit: Is the measurement pipeline sound?",
   "value": "33 findings, 6 P0 (stale CE metric; refusals labelled CE; unchecked X2b padding; unreachable screen bar; Qwen3.5-2B unscreened)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": "review",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: n/a (7 finders + 2 refuters); arm verdict: follow-ups launched",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-043"
   ]
  },
  {
   "id": "F-1909",
   "kind": "fact",
   "claim": "REREAD: Re-read screens in refuse-by-default mode",
   "value": "only 4 visible CE per screen; quote-first is a regression (4→8)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "5 screens",
   "set_status": null,
   "judge": "re-count",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: no (0 calls); arm verdict: frozen verdicts unchanged",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-044"
   ]
  },
  {
   "id": "F-1910",
   "kind": "fact",
   "claim": "GATE17: Groundedness LoRA as a CE detector",
   "value": "0.496 (0.340-0.648)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "57 source-present answers",
   "set_status": null,
   "judge": "AUC vs correctness",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (84cb4fa0); arm verdict: NO (measured the wrong target)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-045"
   ]
  },
  {
   "id": "F-1911",
   "kind": "fact",
   "claim": "LITEPACKS: Do K1-K4 pack sets change the 1.7B?",
   "value": "MAIN no change; self-tests K3 +0.069, K2 +0.045, K1 +0.036; no crowding flag",
   "ci": "K3 +0.069 (+0.038..+0.099)",
   "n": null,
   "seeds": null,
   "set": "MAIN 97; 80 pack-matched self-tests; 100 SE off-pack (dev)",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (frozen + secondary); arm verdict: MAIN NO CHANGE",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-046"
   ]
  },
  {
   "id": "F-1912",
   "kind": "fact",
   "claim": "GATE17-SUPPORT: Does the LoRA separate supported from unsupported answers?",
   "value": "0.965 (0.913-1.000); 0/9 false refusals; 84 % unsupported",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "57 source-present answers; Haiku support labels (κ 0.79)",
   "set_status": null,
   "judge": "AUC",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (98750b05); arm verdict: GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-047"
   ]
  },
  {
   "id": "F-1913",
   "kind": "fact",
   "claim": "RETFLOOR: Remove off-topic appended chunks (V1orC rule)",
   "value": "V1orC removes all 82 off-topic appends, keeps fact reach 98 vs 97; 8.2 pts of K1 coverage were padding",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97 + 180 replays (0 LLM)",
   "set_status": "dev",
   "judge": "counts",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: criterion in arm; arm verdict: rule found; screened later (NO-GO)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-048"
   ]
  },
  {
   "id": "F-1914",
   "kind": "fact",
   "claim": "JUDGE-V2: Binary key-point checklist judge vs judge.v1",
   "value": "CE retest 0.966 vs v1 0.924 (needed 0.974); score SD 0.80x v1 (needed 0.60x)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "3×3 repeats on MAIN 97 control answers",
   "set_status": "dev",
   "judge": "test-retest",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (b7a1b12f); arm verdict: NO-GO; judge.v1 stays",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-049"
   ]
  },
  {
   "id": "F-1915",
   "kind": "fact",
   "claim": "CE-OPUS: Do Haiku CE labels hold under Opus?",
   "value": "CE agree 39/47 (κ 0.63); CE-unwarned 45/47 (κ 0.87); Haiku over-calls CE",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "47 answers",
   "set_status": null,
   "judge": "Claude Haiku vs Claude Opus",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (98a826d9); arm verdict: default-mode metric judge-robust",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-050"
   ]
  },
  {
   "id": "F-1916",
   "kind": "fact",
   "claim": "PACK-HOLES: Do stripped wiki templates leave fact holes?",
   "value": "holes simplewiki 1.29 %, wikivoyage 1.36 %; 3.2 % of K1 v0.1 kept chunks",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "pack files",
   "set_status": null,
   "judge": "scan",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: no (0 LLM scan); arm verdict: converter bug found",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-051"
   ]
  },
  {
   "id": "F-1917",
   "kind": "fact",
   "claim": "REFUSE-UNSUPPORTED: 1.7B + groundedness gate refusing unsupported answers",
   "value": "CE-unwarned 10→4 of 177 (Opus 7→3); false refusals 0/15 (Opus 0/14); coverage 0.435→0.328; projected +9.3 s per answer on Pixel 8a",
   "ci": "U on answered +0.030 (+0.005..+0.057); Opus CE diff CI −0.051..−0.006",
   "n": null,
   "seeds": null,
   "set": "177 = MAIN 97 + 80 pack-matched self-tests",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku; Claude Opus re-judge (CE-OPUS-RU)",
   "date": "2026-10-20",
   "date_text": "2026-10-20/21",
   "source_public": null,
   "caveat": "criterion frozen: yes (af16f29c); arm verdict: PASS (Haiku and Opus); latency blocks shipping",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-052"
   ]
  },
  {
   "id": "F-1918",
   "kind": "fact",
   "claim": "QUANT17: Is Q4_K_M close enough to BF16 for the 1.7B?",
   "value": "Q4 KLD 0.148, top-token 91.2 %; Q8 0.007 / 97.2 %",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "30 prompts, 1,370 tokens",
   "set_status": null,
   "judge": "KLD / top-token",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes; arm verdict: Q4 fails KLD rule → Q8 screen",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-053"
   ]
  },
  {
   "id": "F-1919",
   "kind": "fact",
   "claim": "NULL-CAL / GATE v2: How much do null arms \"fix\" on a screen?",
   "value": "nulls fix 23-31 %; net 6-8; bar = max null net + 2 (K0 10, K1 9)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "new K0 30 / K1 21 problems + 20 sentinels",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (f8239d8b, d1d4b6e4); arm verdict: new screen rule",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-054"
   ]
  },
  {
   "id": "F-1920",
   "kind": "fact",
   "claim": "RETV1ORC: Retrieval hygiene (V1orC) as a quality lever",
   "value": "net 2 (bar 9)",
   "ci": "U on covered −0.026 (−0.064..+0.004)",
   "n": null,
   "seeds": null,
   "set": "GATE v2 K1 screen 41",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (0f2a7016); arm verdict: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-055"
   ]
  },
  {
   "id": "F-1921",
   "kind": "fact",
   "claim": "GATELAT: A cheaper gate layout that keeps quality",
   "value": "context-first + KV reuse 1.8 s but AUC 0.934→0.660",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "124 gated answers",
   "set_status": null,
   "judge": "AUC",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (94e86e40); arm verdict: NO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-056"
   ]
  },
  {
   "id": "F-1922",
   "kind": "fact",
   "claim": "MINISTRAL-3-3B: Ministral-3-3B as the 8 GB model",
   "value": "nets clear bars but false refusals 3-9/20 from dropped [n] citations",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "GATE v2 K0/K1",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (fa1c4a9b); arm verdict: NO-GO (citation format)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-057"
   ]
  },
  {
   "id": "F-1923",
   "kind": "fact",
   "claim": "SHORTCTX2: Short context under GATE v2",
   "value": "net 8 (bar 9); false refusals 6/20",
   "ci": "U cov −0.132 (−0.222..−0.053)",
   "n": null,
   "seeds": null,
   "set": "K1 screen 41",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (97b055a2); arm verdict: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-058"
   ]
  },
  {
   "id": "F-1924",
   "kind": "fact",
   "claim": "GEMMA4-E2B: Gemma 4 E2B as the 8 GB model",
   "value": "screens GO (net 16-21); stage 2 U 0.4055→0.4336",
   "ci": "dU +0.028 (+0.006..+0.052), needed +0.03",
   "n": null,
   "seeds": null,
   "set": "4 GATE v2 screens; MAIN 97 stage 2",
   "set_status": "dev",
   "judge": "Claude Haiku; later Claude Opus",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "criterion frozen: yes (690eeead); arm verdict: stage-2 FAIL by 0.002 (Haiku) | Opus result is , not (Haiku only); cite both.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-059"
   ]
  },
  {
   "id": "F-1925",
   "kind": "fact",
   "claim": "GEMMA4-CONFIRM: Seed 43 and K1 for Gemma",
   "value": "seed 43 +0.030 (+0.007..+0.052); pooled +0.029 (+0.009..+0.048); K1 +0.017 (−0.008..+0.040); gain = fewer refusals",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (399cfe86, measurement); arm verdict: no verdict change",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-060"
   ]
  },
  {
   "id": "F-1926",
   "kind": "fact",
   "claim": "GEMMA-OPUS-MAIN: Gemma vs 1.7B under Opus",
   "value": "s42 U 0.435→0.466; s43 +0.041; pooled +0.036 (+0.011..+0.062)",
   "ci": "s42 +0.031 (+0.005..+0.058)",
   "n": null,
   "seeds": null,
   "set": "MAIN 97",
   "set_status": "dev",
   "judge": "Claude Opus, 362 calls",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (8bf4a4ae); arm verdict: GO by 0.001 (Opus)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-061"
   ]
  },
  {
   "id": "F-1927",
   "kind": "fact",
   "claim": "Q8SCREEN: Q8_0 instead of Q4_K_M",
   "value": "net 9 (bar 10), 7 (bar 9); decode x0.66; +0.73 GB",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "GATE v2 K0/K1",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (22e97896); arm verdict: NO-GO; Q4 stays",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-062"
   ]
  },
  {
   "id": "F-1928",
   "kind": "fact",
   "claim": "QWEN35-2B: Qwen3.5-2B as the 8 GB model",
   "value": "drops [n] citations; false refusals up to 8/20; answers ~2x longer",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "4 screens",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (219ef0e3); arm verdict: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-063"
   ]
  },
  {
   "id": "F-1929",
   "kind": "fact",
   "claim": "OCCDECL: OCC-RAG 1.7B/0.6B with abstain-as-refusal",
   "value": "fixes mostly by refusing; K1 7-9 false refusals",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "4 screens",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (01b04fc8); arm verdict: NO-GO; OCC line closed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-064"
   ]
  },
  {
   "id": "F-1930",
   "kind": "fact",
   "claim": "FOTAPOL-LORA: A public RAG LoRA for Qwen3-1.7B",
   "value": "net 6 (bar 10), 9 (bar 9) with FR 4/20",
   "ci": "U cov −0.084",
   "n": null,
   "seeds": null,
   "set": "2 screens",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (d00e5fa9); arm verdict: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-065"
   ]
  },
  {
   "id": "F-1931",
   "kind": "fact",
   "claim": "CITEFIX: Is the citation-drop failure a format issue?",
   "value": "engine's 50 % key-word support check strips 161 paraphrased citations; format fix flips no gate",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "910 candidate answers",
   "set_status": null,
   "judge": "0 Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (93c85722); arm verdict: open lever: paraphrase-tolerant check",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-066"
   ]
  },
  {
   "id": "F-1932",
   "kind": "fact",
   "claim": "JAN-V1-EDGE: Jan-v1-edge (1.7B SFT+RL) thinking off",
   "value": "K0 net 8 and 7 (bar 10); U_cov fails everywhere",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "4 screens",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (3a1740ea); arm verdict: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-067"
   ]
  },
  {
   "id": "F-1933",
   "kind": "fact",
   "claim": "MIN3CITE: Ministral-3-3B + citation prompt v2",
   "value": "stage 2 Haiku +0.062 (+0.037..+0.089); Opus +0.047 (+0.021..+0.075); ~3x phone time; −0.037 vs 4B (Opus, 61)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "screens; MAIN 97 K0",
   "set_status": "dev",
   "judge": "Claude Haiku; Claude Opus revisit",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (a4d2d9a3; revisit e9813e2c); arm verdict: Haiku PASS; closed as dominated",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-068"
   ]
  },
  {
   "id": "F-1934",
   "kind": "fact",
   "claim": "GEMMA4-PHONE: Gemma on Pixel 8a: speed, memory, quality",
   "value": "73/73, p50 11.8 s, decode 7.0 tok/s, PSS 4.52 GB; phone Gemma 0.452 vs phone 4B 0.535; phone = desktop Gemma (+0.001)",
   "ci": "Gemma − 4B −0.083 (−0.123..−0.047)",
   "n": null,
   "seeds": null,
   "set": "73 phone items; 57 v0.1 parity ids",
   "set_status": "dev",
   "judge": "timing; Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: judge criterion 54cf6c77; arm verdict: measurement",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-069"
   ]
  },
  {
   "id": "F-1935",
   "kind": "fact",
   "claim": "SMOLLM3-3B: SmolLM3-3B",
   "value": "false refusals 3-7/20, all dropped citations",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "4 screens",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (a9f5143d); arm verdict: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-070"
   ]
  },
  {
   "id": "F-1936",
   "kind": "fact",
   "claim": "GRANITE4-1B: granite-4.0-1b",
   "value": "stage 2 U 0.4055→0.4058",
   "ci": "dU +0.0003 (−0.019..+0.020)",
   "n": null,
   "seeds": null,
   "set": "screens; MAIN 97",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (ed62c89a); arm verdict: FAIL (neutral swap)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-071"
   ]
  },
  {
   "id": "F-1937",
   "kind": "fact",
   "claim": "CITESUP: Embedding-cosine rescue of dropped citations",
   "value": "recovers ~1/3 of dropped citations; Ministral still NO-GO",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "2 screens",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (127f17db); arm verdict: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-072"
   ]
  },
  {
   "id": "F-1938",
   "kind": "fact",
   "claim": "GEMMA-VS-4B: How much of the 4B gain does Gemma capture?",
   "value": "U 1.7B 0.406, Gemma 0.434, 4B 0.481; Gemma captures 38 %",
   "ci": "4B−1.7B +0.076 (+0.050..+0.101); 4B−Gemma +0.048 (+0.021..+0.075)",
   "n": null,
   "seeds": null,
   "set": "MAIN 97 K0",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (829e3ba1); arm verdict: measurement",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-073"
   ]
  },
  {
   "id": "F-1939",
   "kind": "fact",
   "claim": "GEMMA-PACKS: Does Gemma read packs better than the 1.7B?",
   "value": "Gemma K3−K0 +0.102 vs 1.7B +0.069",
   "ci": "DiD +0.034 (+0.001..+0.066)",
   "n": null,
   "seeds": null,
   "set": "80 pack-matched self-tests (pack-written)",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (e4c9a4eb); arm verdict: measurement (by a hair)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-074"
   ]
  },
  {
   "id": "F-1940",
   "kind": "fact",
   "claim": "INPACK-SPLIT: Are in-pack items easier; does the matching pack help on real questions?",
   "value": "real matched K3−K0 −0.023 (−0.081..+0.027); self-tests +0.056 (+0.018..+0.096)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97; 17 matched real items; 60 self-tests",
   "set_status": "dev (pack-written self-tests)",
   "judge": "existing Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: post hoc (0 LLM); arm verdict: pack gain only on self-tests",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-075"
   ]
  },
  {
   "id": "F-1941",
   "kind": "fact",
   "claim": "INPACK-STRATUM: Which v0.4 items are in-pack under A∪?",
   "value": "57/75 in-pack under A∪ (55/16 in the 71 headline); 0/75 under the native kit",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.4 tags only (held-out, ids/tags read by script)",
   "set_status": "held-out ruler (v0.4; exposed to two engines 2026-10-24)",
   "judge": "rule",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: rule INPACK-R1 (not frozen then); arm verdict: stratum defined",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-076"
   ]
  },
  {
   "id": "F-1942",
   "kind": "fact",
   "claim": "FACT-REACH: Where do the 20 no-source CEs' key facts live?",
   "value": "top100k 7+2 partial; full enwiki 10+7; published/wave-4 16+4; partly true by construction",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "20 dev items",
   "set_status": "dev",
   "judge": "read-only FTS",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: no (0 LLM); arm verdict: routing/retrieval/reading bind",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-077"
   ]
  },
  {
   "id": "F-1943",
   "kind": "fact",
   "claim": "WIKIVOYAGE-REBUILD: Template whitelist for wikivoyage",
   "value": "fact holes 1.36→0.79 %; literal verdict INVALID by a precondition (27 chunks)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "pack build + 280 replays",
   "set_status": null,
   "judge": "holes",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (1032ab28); arm verdict: candidate, not shipped",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-078"
   ]
  },
  {
   "id": "F-1944",
   "kind": "fact",
   "claim": "Q17B73-PHONE: 1.7B on the same 73 and phone as Gemma",
   "value": "p50 4.2 s, p90 12.2 s, 12.4 tok/s, PSS 3.39 GB",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "73 phone items",
   "set_status": null,
   "judge": "timing",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: no (measurement); arm verdict: measurement",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-079"
   ]
  },
  {
   "id": "F-1945",
   "kind": "fact",
   "claim": "HON-1 WARNCAL: Is the warning line precise?",
   "value": "precision 0.39 (0.32-0.46), recall 0.67 (0.59-0.75); Opus 0.40 / 0.65",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "746 non-decline answers (dev)",
   "set_status": "dev",
   "judge": "existing Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (d641551b); arm verdict: kill not met by a hair",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-080"
   ]
  },
  {
   "id": "F-1946",
   "kind": "fact",
   "claim": "HON-2 SHOWANYWAY: Is \"show anyway\" worth it?",
   "value": "Haiku correct:CE 0.19 (0.12-0.31); Opus 0.54 (0.32-0.83) on pool; MAIN census 11:16 = 0.69",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "492 revealable hidden answers",
   "set_status": null,
   "judge": "Claude Haiku; Claude Opus (WARN-OPUS, REVEAL-CENSUS)",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (same); arm verdict: liability as specified; second confirmation, non-health only AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-081"
   ]
  },
  {
   "id": "F-1947",
   "kind": "fact",
   "claim": "DIVERGENCE-MAP: How many items are deterministic across systems?",
   "value": "30/72 phone v0.1 ids identical text across 1.7B/Gemma/4B; per-item seed r 0.66",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "23 judged conditions",
   "set_status": null,
   "judge": "counts",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: no (0 calls); arm verdict: divergence-first rule",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-082"
   ]
  },
  {
   "id": "F-1948",
   "kind": "fact",
   "claim": "AUNION-PIXEL: Six A∪ packs on Pixel 8a with 1.7B / Gemma / 4B",
   "value": "1.7B answer p50 7.6 vs 4.2 s, PSS 3.41 GB, 0 kills; Gemma PSS 4.38 GB; 4B PSS 4.76 GB; cost = +2.4 s sequential pack search",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "73 phone items",
   "set_status": null,
   "judge": "timing / PSS",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (a852f6bd + addenda); arm verdict: PASS / MARGINAL (no memory cost)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-083"
   ]
  },
  {
   "id": "F-1949",
   "kind": "fact",
   "claim": "FACTPRESENT: Does the self-test pack gain sit on items whose fact the pack adds, and does it hold on real items?",
   "value": "self-test cell a +0.139 (+0.075..+0.201) n 27; real cell a −0.010 (−0.077..+0.050) n 10",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "60 topical self-tests; MAIN 92 keyed",
   "set_status": "dev (pack-written self-tests)",
   "judge": "existing Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (rule frozen first); arm verdict: gain does not reproduce on real items",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-084"
   ]
  },
  {
   "id": "F-1950",
   "kind": "fact",
   "claim": "SPEED-PRECHECK: Are the tiers usable by the ruler's speed rule?",
   "value": "all four USABLE; 1.7B+A∪ p50 T30 4.4 s; 4B p50 T30 20.2 s, complete 26.0 s",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "Pixel 8a traces",
   "set_status": null,
   "judge": "T30 / complete",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (4bd52b3d); arm verdict: 4B also passes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-085"
   ]
  },
  {
   "id": "F-1951",
   "kind": "fact",
   "claim": "FLOOR-ANCHORS: Does judge.v1 give a true zero to declines; is the 1.7B above a closed-book 1B-class model?",
   "value": "bare refusal U 0.058 (0.037-0.081), hedge 0.053, fluent no-content 0.016; 1.5B closed-book 0.259 vs 1.7B 0.294",
   "ci": "1.7B−1.5B +0.035 (−0.019..+0.089)",
   "n": null,
   "seeds": null,
   "set": "30 + 57 v0.1 (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (ed88f377); arm verdict: decline arm INVALID; 1B not separable",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-086"
   ]
  },
  {
   "id": "F-1952",
   "kind": "fact",
   "claim": "TILE-PATH: Can the current-city path be scored on the apps?",
   "value": "8,374 tiles hosted; the native app has no places or location code",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": "code read",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: no (analysis); arm verdict: stratum scored as given",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "E-u-results-087"
   ]
  },
  {
   "id": "F-1953",
   "kind": "fact",
   "claim": "REFVAR: Does reference draw or length change U?",
   "value": "draw A−B moves U 0.015 / 0.008; uncapped vs v1 +0.013 / +0.001; no flip",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "20 v0.1 (dev), 2 uncapped Opus+web draws",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (bc368791); arm verdict: one draw per family suffices",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-088"
   ]
  },
  {
   "id": "F-1954",
   "kind": "fact",
   "claim": "RAG-FUNNEL: Where are facts lost (S1 not in pack → S5)?",
   "value": "K3: S1 45 (39 cooking/money), S2 7, S3 dropped at keep 17, S4 kept-wrong 15, S5 8",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 92 keyed",
   "set_status": "dev",
   "judge": "stage counts",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: no (0 LLM replay); arm verdict: diagnosis",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-089"
   ]
  },
  {
   "id": "F-1955",
   "kind": "fact",
   "claim": "JUDGE-V1_1 / V1_1B: Can one rule line make judge.v1 give declines a true zero?",
   "value": "declines U 0.000, but CE flips 7/40 and 5/40; retest SD 0.0265/0.0295 vs 0.0206",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "decline variants; 40 controls",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (bf560ea9; 6edbacf6); arm verdict: NO-ADOPT both",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-090"
   ]
  },
  {
   "id": "F-1956",
   "kind": "fact",
   "claim": "COVERAGE-MATRIX: Do packs help all three tiers on self-tests; does A∪ harm MAIN?",
   "value": "KA vs K0 +0.052..+0.055 all tiers; 4B−1.7B +0.08 at every pack set; MAIN 1.7B KA−K0 −0.007 (−0.026..+0.012), CE-unwarned 7→12",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "80 self-tests (pack-written); MAIN 97",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (dfc84677); arm verdict: upper bound; no harm shown on U",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-091"
   ]
  },
  {
   "id": "F-1957",
   "kind": "fact",
   "claim": "DRIFT17: Does the 1.7B degrade over a session?",
   "value": "CE-unwarned t1 0/10 vs t4-5 3/20, CI covers 0; one unsafe health error in both arms",
   "ci": "rise CI 0.00..0.30",
   "n": null,
   "seeds": null,
   "set": "longsession-v1 80 (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (5ea4b345); arm verdict: KILLED (no drift detected) AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-092"
   ]
  },
  {
   "id": "F-1958",
   "kind": "fact",
   "claim": "ORACLE-REAL: Does forcing the fact chunk to passage 1 help on real items?",
   "value": "U 0.427 vs 0.412",
   "ci": "+0.015 (−0.029..+0.059)",
   "n": null,
   "seeds": null,
   "set": "30 keyed v0.1 (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (790fe1d0); arm verdict: KILL (reading binds)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-093"
   ]
  },
  {
   "id": "F-1959",
   "kind": "fact",
   "claim": "RV2-SEP-LADDER: Does the ruler separate tiers?",
   "value": "1.5B CB 0.209, 1.7B 0.249, Gemma 0.260, 4B 0.376; coverage DiD fails",
   "ci": "4B−Gemma +0.117 (+0.074..+0.162); 1.7B−1.5B +0.040 (−0.019..+0.096)",
   "n": null,
   "seeds": null,
   "set": "v0.1 51 (dev)",
   "set_status": "dev",
   "judge": "existing Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (53f75edc); arm verdict: 4B separable; 1.7B ≈ closed-book 1.5B (Haiku)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-094"
   ]
  },
  {
   "id": "F-1960",
   "kind": "fact",
   "claim": "RV4-CITY-BOUND: How much can the 7 current-city items swing v0.4?",
   "value": "7 items = 10.7 % of headline weight; swing ≥ 0.107 U",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.4 tags (held-out, script)",
   "set_status": "held-out ruler (v0.4; exposed to two engines 2026-10-24)",
   "judge": "arithmetic",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (53f75edc); arm verdict: KILL branch: separate stratum scored as given",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-095"
   ]
  },
  {
   "id": "F-1961",
   "kind": "fact",
   "claim": "HON-7 WARN-TRANSFER: Does the warning line work for Gemma and the 4B?",
   "value": "Gemma, 1.7B HOLD; 4B precision 0.275 (0.15-0.42) INDETERMINATE",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97",
   "set_status": "dev",
   "judge": "existing Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (f53c867c); arm verdict: refuse-by-default validated for 1.7B/Gemma only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-096"
   ]
  },
  {
   "id": "F-1962",
   "kind": "fact",
   "claim": "HON-6 / CV5-PER100-OPUS: Per 100 questions, refuse-by-default: correct shown vs CE shown by tier",
   "value": "Haiku tie 1.7B=4B 16.5 correct; Opus 4B 22.7, Gemma 23.7, 1.7B 18.6; CE shown 1.0 / 0.0 / 2.1",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97",
   "set_status": "dev",
   "judge": "Claude Haiku; Claude Opus",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (f53c867c; f0206b6f); arm verdict: MIXED (Haiku); tie KILLED (Opus)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-097"
   ]
  },
  {
   "id": "F-1963",
   "kind": "fact",
   "claim": "GEMMA-SAFETY: Is Gemma safe behind the warning stack?",
   "value": "Gemma unsafe 1 vs 1.7B 2",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "65 safety probes",
   "set_status": null,
   "judge": "blinded agent reading",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (1c0389ba); arm verdict: PASS AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-098"
   ]
  },
  {
   "id": "F-1964",
   "kind": "fact",
   "claim": "PK-5 MERGED-PACK: One merged A∪ pack instead of six",
   "value": "kept identity 83.5 % < 95 %; judged U 0.365 vs 0.399",
   "ci": "−0.034 (< −0.02)",
   "n": null,
   "seeds": null,
   "set": "MAIN 97 replay; 16 changed items",
   "set_status": "dev",
   "judge": "identity; Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (59a909f0 + addendum); arm verdict: FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-099"
   ]
  },
  {
   "id": "F-1965",
   "kind": "fact",
   "claim": "PK-3 OVERLAP: Is the self-test pack gain a lexical-overlap artefact?",
   "value": "reweighted to real overlap +0.064 (+0.035..+0.095)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "cand80 vs 37 real golds",
   "set_status": null,
   "judge": "reweighting",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (fe44d05e); arm verdict: KILL (not an artefact; question type/reading)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-100"
   ]
  },
  {
   "id": "F-1966",
   "kind": "fact",
   "claim": "PK-4 PACK-PRUNE: Can any A∪ pack be pruned?",
   "value": "every pack reaches kept context (medlineplus 26 items)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "280-item replay",
   "set_status": null,
   "judge": "kept counts",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (fe44d05e); arm verdict: KILL (keep six)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-101"
   ]
  },
  {
   "id": "F-1967",
   "kind": "fact",
   "claim": "RV1-DECLINE-CU: U from completeness+usefulness only",
   "value": "declines 0, retest ok, but 4B−1.5B gap moves −0.039 (−0.062..−0.015)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.1 51",
   "set_status": "dev",
   "judge": "recompute",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (b0aa1ac3); arm verdict: NO-ADOPT",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-102"
   ]
  },
  {
   "id": "F-1968",
   "kind": "fact",
   "claim": "HON-8 REVEAL-RULE: Is any subset of hidden answers safe to reveal?",
   "value": "none reaches correct:CE ≥ 1.0; best 0.37",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "492 revealable rows",
   "set_status": null,
   "judge": "counts",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (b0aa1ac3); arm verdict: KILL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-103"
   ]
  },
  {
   "id": "F-1969",
   "kind": "fact",
   "claim": "RD-1 CONVERSION-MAP: Does chunk form explain why the 1.7B ignores a present fact?",
   "value": "conversion: 1.7B 0.33, Gemma 0.72, 4B 0.57; best ceiling 0.002 U",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "fact-present rows",
   "set_status": null,
   "judge": "existing Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (84c2c5a2); arm verdict: KILL (reading capacity)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-104"
   ]
  },
  {
   "id": "F-1970",
   "kind": "fact",
   "claim": "CHUNKCLEAN: Hand-cleaned fact sentence vs raw chunk",
   "value": "C−R +0.031",
   "ci": "+0.031 (−0.009..+0.073)",
   "n": null,
   "seeds": null,
   "set": "30 real v0.1",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (10f3c2df); arm verdict: INCONCLUSIVE",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-105"
   ]
  },
  {
   "id": "F-1971",
   "kind": "fact",
   "claim": "CV5-LADDER-OPUS: Ladder under Opus",
   "value": "4B−1.7B +0.109 (+0.063..+0.160); 4B−Gemma +0.098; 1.7B−1.5B CB +0.090 (+0.009..+0.170) vs Haiku +0.038",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.1 51",
   "set_status": "dev",
   "judge": "Claude Opus, 324 calls",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (3870ac70); arm verdict: 4B gap holds; \"1.7B level with 1B-class\" sentence killed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-106"
   ]
  },
  {
   "id": "F-1972",
   "kind": "fact",
   "claim": "RD-4 DILUTION-K: Gold-only context vs gold + rest",
   "value": "k1 − R −0.009",
   "ci": "−0.009 (−0.049..+0.033)",
   "n": null,
   "seeds": null,
   "set": "30 ORACLE-REAL items",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (af00dcaa); arm verdict: KILL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-107"
   ]
  },
  {
   "id": "F-1973",
   "kind": "fact",
   "claim": "CV5-RECOMPUTE / EXPOSURE-AUDIT: Do page numbers re-derive; are v0.4 and held-out ledgers clean?",
   "value": "263/270 re-derive (7 mismatches, 2.6 %); 25/295 unsourced; v0.4 0 run files; 5 missing ledger rows",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "295 page numbers; 2,203 run files",
   "set_status": null,
   "judge": "script",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (78f43da6); arm verdict: PASS with one label fix",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-108"
   ]
  },
  {
   "id": "F-1974",
   "kind": "fact",
   "claim": "WARN-TUNE: Can a trace-feature rule lift warning precision to 0.55?",
   "value": "goes the wrong way on MAIN (0.407→0.383)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "fit 180, test MAIN 97",
   "set_status": "dev",
   "judge": "existing labels",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (1799c624); arm verdict: KILL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-109"
   ]
  },
  {
   "id": "F-1975",
   "kind": "fact",
   "claim": "FEWSHOT-READ: Two worked examples in the 1.7B prompt",
   "value": "U 0.396→0.395",
   "ci": "−0.001 (−0.015..+0.014)",
   "n": null,
   "seeds": null,
   "set": "stage1-pool 87",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (081c410c); arm verdict: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-110"
   ]
  },
  {
   "id": "F-1976",
   "kind": "fact",
   "claim": "SAMPLE2-GATE: Best-of-2 seeds picked by the gate",
   "value": "decode x2",
   "ci": "+0.003 (−0.007..+0.014)",
   "n": null,
   "seeds": null,
   "set": "stage1-pool 87",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (081c410c); arm verdict: NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-111"
   ]
  },
  {
   "id": "F-1977",
   "kind": "fact",
   "claim": "HAIKU-CAL: Map Haiku-low to Opus on double-labelled units",
   "value": "κ CE 0.75, correct 0.49 (warned 0.22); U shift Opus−Haiku +0.022",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "1,911 units",
   "set_status": null,
   "judge": "κ",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (spec frozen first); arm verdict: calibration map",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-112"
   ]
  },
  {
   "id": "F-1978",
   "kind": "fact",
   "claim": "DECLINE2STAGE: A Haiku stage-1 non-answer classifier before judge.v1",
   "value": "recall 90/90; bar (b) 2 of 49 zeroed (both the engine refusal string) vs cap 1",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "FLOOR declines 90; MAIN 49; WARN 27",
   "set_status": "dev",
   "judge": "Claude Haiku, 258 calls",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (43d52938); arm verdict: FAIL by letter → RULING R1 adopted",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-113"
   ]
  },
  {
   "id": "F-1979",
   "kind": "fact",
   "claim": "X2B-RETEST / X2B-OFF-MAIN: Is the X2b head pass worth its tokens on real items?",
   "value": "real +0.033 (−0.100..+0.173) n 25; OFF pooled dU_d +0.026 (−0.020..+0.073), CE-unwarned +2; prompt −42 %",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "stage1-pool 87; MAIN 77 live, 2 seeds",
   "set_status": "dev",
   "judge": "Claude Haiku; Claude Opus on flips",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (3a3aaf0b; 0415e6eb + addendum); arm verdict: underpowered; OFF safe only at edge (kit-only)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-114"
   ]
  },
  {
   "id": "F-1980",
   "kind": "fact",
   "claim": "HYBRID-READER: Route hard items to the 4B reader by retrieval features",
   "value": "routes 5/97, +0.004; random routing similar",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97",
   "set_status": "dev",
   "judge": "existing Claude Haiku/Claude Opus",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (87060c26); arm verdict: KILL (chance level)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-115"
   ]
  },
  {
   "id": "F-1981",
   "kind": "fact",
   "claim": "ROUTING-R1: Fix routing losses",
   "value": "7/97 losses; fixes drop safety probes off the health path (up to 32/64)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97 + 64 probes",
   "set_status": "dev",
   "judge": "replay + Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (0fff7357); arm verdict: none adopted AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-116"
   ]
  },
  {
   "id": "F-1982",
   "kind": "fact",
   "claim": "KEEP-K1 / KEEP-E: Keep-step variants to bring dropped fact chunks in",
   "value": "best 7/17 at +50 % tokens; KEEP-E net U −0.010",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97 + 180 replays; 14 changed",
   "set_status": "dev",
   "judge": "0 LLM; Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (keep-e 4421544c); arm verdict: CLOSED / KILL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-117"
   ]
  },
  {
   "id": "F-1983",
   "kind": "fact",
   "claim": "E1-PARTIME / E1-PAR: Parallel pack search on the bench engine",
   "value": "projection 7.57→5.5-6.0 s; device: search 3.89→1.93 s, answer 10.32→7.39 s; cache-64 arm KILL by rule",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "phone traces; 10 phone items per arm",
   "set_status": null,
   "judge": "timing",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (ffd0d931); arm verdict: Promise.all PAYS (bench engine only)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-118"
   ]
  },
  {
   "id": "F-1984",
   "kind": "fact",
   "claim": "NOTE2-TRANSFER: Do desktop hide rates transfer to the phone?",
   "value": "1.7B A∪ phone vs desktop +26 pp",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "57 shared v0.1",
   "set_status": "dev",
   "judge": "counts",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (b6b1f787); arm verdict: KILLED",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-119"
   ]
  },
  {
   "id": "F-1985",
   "kind": "fact",
   "claim": "BLOCKCACHE: Decompressed-block LRU for pack search",
   "value": "256 slots −78 % decompressions but +65 MB > 32 MB",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "session replay",
   "set_status": null,
   "judge": "counts",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (4dab635a); arm verdict: CLOSED as frozen",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-120"
   ]
  },
  {
   "id": "F-1986",
   "kind": "fact",
   "claim": "DEFAULT-SET: Which A∪ packs to install by default",
   "value": "frozen clause keeps visa and medlineplus; options table for author",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "112 ids + 64 probes",
   "set_status": null,
   "judge": "replay",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (2c8e3010); arm verdict: KEEP",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-121"
   ]
  },
  {
   "id": "F-1987",
   "kind": "fact",
   "claim": "C1-REASON: Why are 16 owned-pack facts missed?",
   "value": "7 retrieval, 4 routing, 2 build, 2 install-only, 1 reading; install reaches 2/9",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "16 facts",
   "set_status": null,
   "judge": "replay",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: no (0 LLM); arm verdict: C1 closes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-122"
   ]
  },
  {
   "id": "F-1988",
   "kind": "fact",
   "claim": "FA-CARDS: Are there coverable first-aid card misses?",
   "value": "0 beyond epinephrine; trigger edit routes 0/2 epi turns",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "245 rows",
   "set_status": null,
   "judge": "replay",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (42b910a3); arm verdict: KILLED AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-123"
   ]
  },
  {
   "id": "F-1989",
   "kind": "fact",
   "claim": "CONVERT-FIX: Converter drops unit-conversion values",
   "value": "artefact chunks 3,176→524; simplewiki r2 published 15:01",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "pack rebuild",
   "set_status": null,
   "judge": "counts",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: n/a (bug + 11 tests); arm verdict: shipped (pack r2)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-124"
   ]
  },
  {
   "id": "F-1990",
   "kind": "fact",
   "claim": "ZSTD-BENCH / PARTIAL-DECOMP / REPACK16: Faster pack block decode",
   "value": "WASM zstd 3-4x faster (no WASM on Hermes, unresolved); 16 KB repack −18 % bytes at 64 slots (rule 50 %)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "200 blocks; replays",
   "set_status": null,
   "judge": "timing",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (REPACK16 daedb7dd); arm verdict: REPACK16 KILL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-125"
   ]
  },
  {
   "id": "F-1992",
   "kind": "fact",
   "claim": "A1-SUCCINCT17: Time saved by Succinct for small models",
   "value": "Detailed +348 tokens p50 → ~24 s per answer at 14.3 tok/s (projection)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "56 v0.1 native exports",
   "set_status": "dev",
   "judge": "tokens × tok/s",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (d037c010); arm verdict: SUPPORTED pending device",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "E-u-results-127"
   ]
  },
  {
   "id": "F-1993",
   "kind": "fact",
   "claim": "A4-CAP512: Is a 512-token Detailed cap a lever?",
   "value": "saves 0 s p50; truncates 10.9 %",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "404 Detailed answers",
   "set_status": null,
   "judge": "tokens",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (a1d6bf6e); arm verdict: NOT A LEVER",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-128"
   ]
  },
  {
   "id": "F-1994",
   "kind": "fact",
   "claim": "LENGTH-SLOPE: Does judge.v1 reward length?",
   "value": "no length reward except the 1.7B under Opus (+0.31 per doubling)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "existing rows",
   "set_status": null,
   "judge": "within-item slope",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: no (0 LLM, 13,310 rows); arm verdict: Succinct not penalised at ≥4B",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-129"
   ]
  },
  {
   "id": "F-1995",
   "kind": "fact",
   "claim": "G1-SWA5: Gemma 4 sliding-window KV bug in 5-turn sessions on the phone runtime",
   "value": "0 decode/KV markers, 0 garbled",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "10 turns per arm, Pixel 8a",
   "set_status": null,
   "judge": "markers",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (b89f1f1a); arm verdict: KILLED (no bug seen)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-130"
   ]
  },
  {
   "id": "F-1997",
   "kind": "fact",
   "claim": "SHIP15B-PACKS: Do packs help the 1.5B the app ships?",
   "value": "0.270 vs 0.255; CE shown 25.5 vs 31.4 per 100",
   "ci": "+0.015 (−0.036..+0.068)",
   "n": null,
   "seeds": null,
   "set": "v0.1 51",
   "set_status": "dev",
   "judge": "Claude Opus + D2S; Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (4f9e7d5a); arm verdict: NO-GAIN",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-132"
   ]
  },
  {
   "id": "F-1998",
   "kind": "fact",
   "claim": "STYLE15B: Succinct vs Detailed on 1.5B and 1.7B",
   "value": "1.5B 0.220 vs 0.133; 1.7B 0.263 vs 0.224; CE shown halves",
   "ci": "1.5B +0.087 (+0.038..+0.143); 1.7B +0.038 (−0.005..+0.084)",
   "n": null,
   "seeds": null,
   "set": "v0.1 51",
   "set_status": "dev",
   "judge": "Claude Haiku; Claude Opus diagnostic 11",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (474ac0a0); arm verdict: Succinct NONINFERIOR",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-133"
   ]
  },
  {
   "id": "F-2000",
   "kind": "fact",
   "claim": "CB17: Do packs move the 1.7B over closed-book?",
   "value": "CB 0.337 vs K0 0.351 (Opus 51); 63/97 texts identical",
   "ci": "K0−CB +0.015 (−0.035..+0.056)",
   "n": null,
   "seeds": null,
   "set": "MAIN 97; v0.1 51",
   "set_status": "dev",
   "judge": "Claude Opus 58; Claude Haiku 291",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: criterion in arm; arm verdict: packs not separable on aggregate",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-135"
   ]
  },
  {
   "id": "F-2001",
   "kind": "fact",
   "claim": "LA1 / LA5: Can section 0 state app and engine numbers without misattribution?",
   "value": "LA1 0 misattribution; LA5 withhold = material omission",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "text variants",
   "set_status": null,
   "judge": "1 Claude Opus reviewer + 3 cold readers",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (49d1c5ac; 304baaf0); arm verdict: LA1 PASS; LA5 KILLED",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-136"
   ]
  },
  {
   "id": "F-2002",
   "kind": "fact",
   "claim": "C1-GAP-LEDGER: Sourced cause ledger for the app gap",
   "value": "5 rows verified (no 1.7B in catalog; Detailed hook; 0 packs; no warning text; KB cutoff 0.7); layers not established",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "code + device files",
   "set_status": "dev",
   "judge": "checker",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: ledger sha 3e233f6e; arm verdict: ledger",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-137"
   ]
  },
  {
   "id": "F-2006",
   "kind": "fact",
   "claim": "NOREPACK97: Cross-host determinism flag",
   "value": "97/97 identical across hosts",
   "ci": "dU −0.012 (−0.030..+0.005)",
   "n": null,
   "seeds": null,
   "set": "MAIN 97",
   "set_status": "dev",
   "judge": "identity; Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (27d05cde); arm verdict: adopted for pairing only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-141"
   ]
  },
  {
   "id": "F-2007",
   "kind": "fact",
   "claim": "RC1-STAGE1: Does stage 1 wrongly zero long app answers?",
   "value": "0 of 51 marked non-answer",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "51 RC1 answers",
   "set_status": null,
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (824d94f5); arm verdict: PASS (low power)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "E-u-results-142"
   ]
  },
  {
   "id": "F-2008",
   "kind": "fact",
   "claim": "FACT-AVAIL: Which v0.4 items have their answer in the A∪ packs?",
   "value": "AVAILABLE 19 / PARTIAL 25 / NONE 27 (strict 15/23/33); validation 34/40",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.4 71 headline (held-out; labels quarantined)",
   "set_status": "held-out ruler (v0.4; exposed to two engines 2026-10-24)",
   "judge": "Claude Haiku 329 + Claude Opus 108 labels",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (1ac3fa96 + addendum); arm verdict: label set (quarantined)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "E-u-results-143"
   ]
  },
  {
   "id": "F-2009",
   "kind": "fact",
   "claim": "DECODE-GROUND: Grounding at the decoder (grammar select, context-aware decoding, closed-book fallback)",
   "value": "all arms net 4-6 (bar 9)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "GATE v2 K1 screen; MAIN 97",
   "set_status": "dev",
   "judge": "Claude Haiku 276",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (92d3340a; SA 19956cfe); arm verdict: CLOSED (null)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-144"
   ]
  },
  {
   "id": "F-2010",
   "kind": "fact",
   "claim": "QFORM: Does question phrasing decide whether the pack chunk is found, and how well does the 1.7B answer when it is?",
   "value": "hit@kept base 50 %, entity-free 10 %, follow-up 8 %, situational 25 %; reference kept U 0.88 (CE 2/43) vs not 0.19 (CE 75/147)",
   "ci": "F1 −40 (−54..−23 pts), F8 −42, F3 −25",
   "n": null,
   "seeds": null,
   "set": "48 pack facts × 9 forms = 432 (dev, agent-written)",
   "set_status": "dev",
   "judge": "replay; Claude Haiku 384",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: yes (8371f509); arm verdict: form is a retrieval effect",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-145"
   ]
  },
  {
   "id": "F-2011",
   "kind": "fact",
   "claim": "QFORM-W74: Previous turn + fragment as the search query",
   "value": "kept 4/48 → 24/48; the app's own follow-up rule already gives 20/48",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "48 follow-up fragments",
   "set_status": null,
   "judge": "replay",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "criterion frozen: no (0 LLM); arm verdict: bench-engine artefact",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-146"
   ]
  },
  {
   "id": "F-2012",
   "kind": "fact",
   "claim": "RETRIEVE-SWEEP: Can engine knobs guarantee the chunk is found?",
   "value": "best sibling arm +7.9 pts (junk 2.4x); real recovery ≤ 2/25",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "QFORM 432; funnel 25/22",
   "set_status": null,
   "judge": "replay",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: yes (94cde143); arm verdict: no arm passes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-147"
   ]
  },
  {
   "id": "F-2013",
   "kind": "fact",
   "claim": "ARM-G chunk vectors: Chunk-level vectors for packs",
   "value": "kept 21→32-36 %; answers +0.109 (+0.026..+0.190), final +0.125 (+0.049..+0.204) on pack items; whole kit real recovery 3/25",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "216 QFORM items; MAIN 97",
   "set_status": "dev",
   "judge": "replay; Claude Haiku 336-400",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: criterion in arm (G2 dcac6f4b); arm verdict: pack items yes; G2 FAIL on real items",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-148"
   ]
  },
  {
   "id": "F-2014",
   "kind": "fact",
   "claim": "T1-TERMS: Model-free search-term extraction for situational questions",
   "value": "situational 25 %→23 %",
   "ci": "−2 (−15..+10)",
   "n": null,
   "seeds": null,
   "set": "QFORM 432",
   "set_status": null,
   "judge": "replay",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: yes (b7d73bf0); arm verdict: FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-149"
   ]
  },
  {
   "id": "F-2015",
   "kind": "fact",
   "claim": "COVERAGE-MAP: Are missing dev facts absent from packs on disk?",
   "value": "installing 18 packs moves S1 52→4 but kept only 23→28",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 92 keyed",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: no (0 LLM); arm verdict: after install the gap is retrieval",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-150"
   ]
  },
  {
   "id": "F-2016",
   "kind": "fact",
   "claim": "SEARCH-AUTOPSY / H1-H3: Autopsy of 25 lost real items and 3 targeted fixes",
   "value": "only ~9 are retrieval losses; H1+H2 answer screen −0.023 (−0.052..+0.006)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97",
   "set_status": "dev",
   "judge": "replay; Claude Haiku 150",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: yes (94cde143 rule); arm verdict: FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-151"
   ]
  },
  {
   "id": "F-2017",
   "kind": "fact",
   "claim": "B3-FULLWIKI: Do full-Wikipedia shards deliver facts?",
   "value": "each shard covers 1-5 more facts, kept −3..+3",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97 × 15 shards",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: criterion in arm; arm verdict: COVERAGE-ONLY (scored with old merge)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-152"
   ]
  },
  {
   "id": "F-2018",
   "kind": "fact",
   "claim": "H4-MERGE (gidf): Global-IDF re-scoring across packs",
   "value": "+6.2 pts kept (27 gained, 0 lost, junk 1.10x); rrf worse",
   "ci": "+6.2 (+3.9..+8.6)",
   "n": null,
   "seeds": null,
   "set": "QFORM six separate packs; MAIN 92",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: criterion in arm; arm verdict: first no-harm composition gain",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-153"
   ]
  },
  {
   "id": "F-2019",
   "kind": "fact",
   "claim": "COMPOSE: Do 10-13 small packs compose?",
   "value": "delivered 22→29 (stock) / 23→32 (gidf), crowding flat; answers gidf +0.009 (−0.022..+0.039)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 92/97",
   "set_status": "dev",
   "judge": "replay; Claude Haiku 204",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: yes (2d071770); arm verdict: delivery yes; answers flat",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-154"
   ]
  },
  {
   "id": "F-2020",
   "kind": "fact",
   "claim": "PRECISION-GATE: Withhold wrong context (score gate / select gate)",
   "value": "P1 +0.016 (−0.047..+0.076) n 16; P2 +0.004",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97",
   "set_status": "dev",
   "judge": "Claude Haiku 38",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: yes (a350a88b); arm verdict: FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-155"
   ]
  },
  {
   "id": "F-2021",
   "kind": "fact",
   "claim": "AUTOPSY52 / H4b: Why 52 pack facts are not delivered; global avgdl",
   "value": "SE packs lose on per-pack df cut, no chunk vectors, health misfire; gidf2 −1.4 vs gidf",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 92",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: no / criterion; arm verdict: H4b NO AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-156"
   ]
  },
  {
   "id": "F-2022",
   "kind": "fact",
   "claim": "F1b: Chunk vectors for the cooking/money Q&A packs",
   "value": "delivered 32→45 (article level; chunk level 32); answers −0.026 (−0.051..−0.002)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 92/97",
   "set_status": "dev",
   "judge": "replay; Claude Haiku 132",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: yes (SEFIX 07ea2b5e); arm verdict: delivers; answers FAIL",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-157"
   ]
  },
  {
   "id": "F-2023",
   "kind": "fact",
   "claim": "SEFIX-ALL: Lexical fixes F1-F4 for Q&A packs + gidf",
   "value": "chunk-level 28→47; pooled +0.023 (+0.001..+0.048); Opus CV5 −0.002 (−0.062..+0.061), CE halves; Opus where fact delivered (cm) +0.173 (+0.082..+0.266)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97, 2 seeds; Opus CV5 42; Opus cm 22",
   "set_status": "dev",
   "judge": "Claude Haiku; Claude Opus",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: yes (07ea2b5e); arm verdict: LEVER-CANDIDATE, then withdrawn (F3 unsafe) AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-158"
   ]
  },
  {
   "id": "F-2024",
   "kind": "fact",
   "claim": "F3-SAFETY / F3b: Is the F3 health no-source fallback safe?",
   "value": "F3 fires 21/65, unsafe 9→13; F3b-A vacuous, F3b-B unsafe 9→12",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "65 safety probes",
   "set_status": null,
   "judge": "blinded agent reading",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: yes (15be6d08; 458ee048); arm verdict: FAIL; F3 withdrawn AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-159"
   ]
  },
  {
   "id": "F-2025",
   "kind": "fact",
   "claim": "SEFIX-124: F1/F2/F4 without F3",
   "value": "chunk-level 42/92, 0 lost; pooled +0.013",
   "ci": "+0.013 (−0.008..+0.035)",
   "n": null,
   "seeds": null,
   "set": "MAIN 97, 2 seeds",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: yes (07ea2b5e); arm verdict: not a lever-candidate; adopted for delivery",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-160"
   ]
  },
  {
   "id": "F-2026",
   "kind": "fact",
   "claim": "SEARCHPAR R1-R5: Dense re-rank, header vectors, BM25 k1/b, near-dup drop, sticky sources",
   "value": "R1 mix +10 chunk-level facts; R2/R3/R4 NO; R5 subject carry +17/48 F8",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 92; QFORM; F8 pairs",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: yes (d8fca0ee); arm verdict: R1 candidate; others NO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-161"
   ]
  },
  {
   "id": "F-2027",
   "kind": "fact",
   "claim": "STACK: SEFIX + chunk vectors + R1 together",
   "value": "chunk-level 45 vs bar 52; routes are alternatives, not a stack",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 92",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: yes (c0fabf99); arm verdict: NO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-162"
   ]
  },
  {
   "id": "F-2028",
   "kind": "fact",
   "claim": "READER / COMBINED: Reader fixes (order, cite-or-ground, keep answers); SEFIX + fact-first",
   "value": "none READS; 11/44 \"delivered\" had no fact in kept text; COMBINED +0.022 vs K3, fact-first hurts v0.1",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "44 delivered ids; MAIN 97 × 2 seeds",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: yes (f361f738; 9a9c8550); arm verdict: NO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-163"
   ]
  },
  {
   "id": "F-2029",
   "kind": "fact",
   "claim": "CHUNKLEVEL: Re-score delivery at chunk level",
   "value": "F1b 45→32; SEFIX-ALL 55→47",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "42 replays",
   "set_status": null,
   "judge": "counts",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "criterion frozen: n/a (0 LLM); arm verdict: correction",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-164"
   ]
  },
  {
   "id": "F-2030",
   "kind": "fact",
   "claim": "DETECTOR-HN / FT / FT2: Retrain the health detector against cooking/money misfires",
   "value": "FP 3.4 %→0 %, but recall losses 76-185 positives and probe routes 47→13-40",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "held-out titles; probes; MAIN",
   "set_status": "held-out",
   "judge": "0 LLM reads",
   "date": "2026-10-22",
   "date_text": "2026-10-22/23",
   "source_public": null,
   "caveat": "criterion frozen: yes (7b7bed00; 53002568; 34301679); arm verdict: NOT ACCEPTED (all) AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-165"
   ]
  },
  {
   "id": "F-2031",
   "kind": "fact",
   "claim": "NIGHT-A-KIT: Do more crypto packs add v0.1 facts?",
   "value": "28 = 28",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.1 52; MAIN 92",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (530025688); arm verdict: not adopted",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-166"
   ]
  },
  {
   "id": "F-2032",
   "kind": "fact",
   "claim": "NIGHT-B-TABLE: On real items, are delivered-fact answers better than non-delivered?",
   "value": "delivered 0.420 vs not 0.419; within-item arm−control +0.044 (+0.009..+0.082) n 42; Opus slice +0.013 (−0.044..+0.072)",
   "ci": "between items +0.001 (−0.071..+0.069)",
   "n": null,
   "seeds": null,
   "set": "MAIN 92 keyed, 2 seeds; Opus CV5 28",
   "set_status": "dev",
   "judge": "Claude Haiku; Claude Opus",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (530025688); arm verdict: NOT PUBLISHABLE",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-167"
   ]
  },
  {
   "id": "F-2033",
   "kind": "fact",
   "claim": "SUPPORT-FLAG 2x2: Same with a judged support flag",
   "value": "supported vs not 0.44 vs 0.43; within-item +0.040 (−0.008..+0.087)",
   "ci": "gap +0.005 pooled",
   "n": null,
   "seeds": null,
   "set": "MAIN 97",
   "set_status": "dev",
   "judge": "Claude Haiku 190",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (93f342bd); arm verdict: claim \"delivered facts are answered well\" does not hold between items",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-168"
   ]
  },
  {
   "id": "F-2034",
   "kind": "fact",
   "claim": "W98: Cooking/money-source veto on the health route",
   "value": "silences 5/5 real health probes",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "65 probes; MAIN 97",
   "set_status": "dev",
   "judge": "counts",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: no (traces); arm verdict: NO AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-169"
   ]
  },
  {
   "id": "F-2036",
   "kind": "fact",
   "claim": "ENGFIX-1023 / CO-PTES / PLUGFIX: Engine fixes (CO vehicles, digit runs, seed cue, PT/ES CO, plugin routes)",
   "value": "digits 3.7 s→14 ms; route 89→103/110; suite 1,645 green",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "110 plugin rows; suites",
   "set_status": null,
   "judge": "replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: n/a (tests); arm verdict: patches staged",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-171"
   ]
  },
  {
   "id": "F-2037",
   "kind": "fact",
   "claim": "TOPICS27 / SUCCINCT: Are the app's 27 suggested topics answered with support?",
   "value": "unsupported Detailed 0.468 vs Succinct 0.283; invented specifics 98 vs 7; KEEP 7→19/27",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "27 topics with Vital pack (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku sentence audit",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (5f8ba772; 84835bc8); arm verdict: Succinct for 1.5-1.7B",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-172"
   ]
  },
  {
   "id": "F-2038",
   "kind": "fact",
   "claim": "ENG-BANNER-KEPT / UXT-135: Why does the \"Not found\" line show when passages were kept?",
   "value": "false banner 37/97; patch switches 37/38 to \"passages found don't support\"",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "TOPICS27; MAIN 97",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: n/a; arm verdict: patch ready; recount closes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-173"
   ]
  },
  {
   "id": "F-2039",
   "kind": "fact",
   "claim": "VITAL-MISSES / KEEPPIN / PIN2 / FMT1-SCAN: Recover the 4 missed topics",
   "value": "4 losses at 4 stages; pins recover Kyoto but fail context clause; PIN2 screen thin +0.018 (−0.034..+0.068)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "27 topics; MAIN 97",
   "set_status": "dev",
   "judge": "replay; PIN2 Claude Haiku 24",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (80316ed8; 83bbae38; 7b08c21a; 79fefb9c); arm verdict: NO / HOLD",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-174"
   ]
  },
  {
   "id": "F-2040",
   "kind": "fact",
   "claim": "W78 (BINARY / B / C / PACKFMT): Cut phone reading cost of the dense stage",
   "value": "kit 125.09→21.26 MB per question; cold stage 1 620→92 ms; kept 96/97; facts 42/42; W78B NO",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97; QFORM 216",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (e9d5e44a; e3279284; f26642ca; 7a81fd33); arm verdict: ADOPT-CANDIDATE (phone unmeasured)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-175"
   ]
  },
  {
   "id": "F-2041",
   "kind": "fact",
   "claim": "PARA135: 1/3/5 paragraphs on the 1.7B",
   "value": "unsupported 0.33/0.50/0.52; U 0.228/0.233/0.218; CE 37/43/49 per 100",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "27 topics; v0.1 51",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (e59b069a); arm verdict: Succinct best; support breaks at paragraph 2",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-176"
   ]
  },
  {
   "id": "F-2042",
   "kind": "fact",
   "claim": "DEEPMODE: Deep Research multipass on the 1.7B",
   "value": "U 0.174 vs Detailed 0.224 vs Succinct 0.263; 2.6-5.6x slower",
   "ci": "vs Detailed −0.050 (−0.087..−0.015)",
   "n": null,
   "seeds": null,
   "set": "27 topics; v0.1 51",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (8753e6b7); arm verdict: worse; hide",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-177"
   ]
  },
  {
   "id": "F-2043",
   "kind": "fact",
   "claim": "UXT-24: Personal files above Vital; follow-up source keep",
   "value": "fact kept 10/14 (Vital on) vs 13/14 off; rescue 14/14",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "29 questions on the app's JVM harness",
   "set_status": null,
   "judge": "replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (c2faa6eb); arm verdict: YES-WITH-GUARD",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-178"
   ]
  },
  {
   "id": "F-2044",
   "kind": "fact",
   "claim": "STYLE-SEFIX: Does Detailed use delivered facts better?",
   "value": "dU −0.028; delivered −0.047; CE 25.8→46.4 per 100",
   "ci": "−0.028 (−0.048..−0.008)",
   "n": null,
   "seeds": null,
   "set": "MAIN 97",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (25478a97); arm verdict: Succinct stays",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-179"
   ]
  },
  {
   "id": "F-2045",
   "kind": "fact",
   "claim": "PLUGBATCH-SELF: Plugin-stress routes on our 1.7B engine",
   "value": "85/110; engfix 89; after PLUGFIX 103",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "110 public rows",
   "set_status": "dev",
   "judge": "route replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (8dfa7a8a); arm verdict: measurement",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "E-u-results-180"
   ]
  },
  {
   "id": "F-2047",
   "kind": "fact",
   "claim": "U5-FOLLOWUP-PACKS: Does the turn-1 source survive the follow-up with the pack on?",
   "value": "survives 1/19 and 4/13; subject carry recovers 15/18 at 1.22x off-subject",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "48 F8 pairs; 20 long-session turns",
   "set_status": null,
   "judge": "replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (fab6359f); arm verdict: no fix by the line",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-182"
   ]
  },
  {
   "id": "F-2048",
   "kind": "fact",
   "claim": "UXMATRIX: In each mode, is the answer right when the pack holds the content, and what happens without it?",
   "value": "Succinct ON CWS 0.28 vs OFF 0.03; Detailed 0.18 vs 0.03; correct 35 vs 27 % (ON), 28 vs 17 % (OFF, approx.)",
   "ci": "ON−OFF CWS +0.27 (Succinct)",
   "n": null,
   "seeds": null,
   "set": "60 items (40 cm + 20 v0.1)",
   "set_status": "dev",
   "judge": "Claude Haiku outcome labels",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (0cc3ca20); arm verdict: Succinct beats Detailed Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-183"
   ]
  },
  {
   "id": "F-2049",
   "kind": "fact",
   "claim": "WARN-KEPT: Warn cited-but-unsupported answers",
   "value": "warns 9/26 wrong at cost 5/37 correct",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "UXMATRIX 120; MAIN",
   "set_status": "dev",
   "judge": "0 LLM",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (d5b5436d); arm verdict: NO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-184"
   ]
  },
  {
   "id": "F-2050",
   "kind": "fact",
   "claim": "PACKPROOF: With vs without the pack on its own self-tests",
   "value": "correct-any 0.48 vs 0.26; when own article kept 0.65 vs 0.28; OFF wrong-confident 0.17",
   "ci": "+0.23 (+0.16..+0.31)",
   "n": null,
   "seeds": null,
   "set": "180 self-tests of 9 packs (pack-written)",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (37f74d6f); arm verdict: P1 MISS (bar 0.75); honest-when-not via warning Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-185"
   ]
  },
  {
   "id": "F-2051",
   "kind": "fact",
   "claim": "PACK-GAP: Would new packs close the undelivered gap?",
   "value": "44/50 already in K13 (search problem); new packs ≤ ~1 fact each",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "50 undelivered MAIN items",
   "set_status": "dev",
   "judge": "FTS probe",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: no (0 LLM); arm verdict: no new packs",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-186"
   ]
  },
  {
   "id": "F-2052",
   "kind": "fact",
   "claim": "NOWIKI: Kit without simplewiki and Vital",
   "value": "MAIN delivered 42→33",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97; 160 self-tests",
   "set_status": "dev (pack-written self-tests)",
   "judge": "replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (5d40edcd); arm verdict: KEEP general packs",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-187"
   ]
  },
  {
   "id": "F-2053",
   "kind": "fact",
   "claim": "QREWRITE: Ceiling of question rewriting (oracle)",
   "value": "MAIN 42→51 (+9 < +10); ST kept 100→118 (+18 < +20)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97; 180 self-tests",
   "set_status": "dev (pack-written self-tests)",
   "judge": "replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (60fdc688); arm verdict: NO-CEILING (stopped at stage 0)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-188"
   ]
  },
  {
   "id": "F-2054",
   "kind": "fact",
   "claim": "PACKSHAPE / V2-CONFIRM: Factbook as short country-named sentences",
   "value": "fact chunk kept 7→15/20; correct 5→10/20 (pooled 0.50); Opus U 0.316→0.436",
   "ci": "Opus +0.121 (+0.030..+0.221)",
   "n": null,
   "seeds": null,
   "set": "20 factbook self-tests (pack-written), 2 seeds",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku; Claude Opus",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (89f565ca; c5fb7b8b); arm verdict: LEVER-CANDIDATE; NOT-CONFIRMED by 1 new warned CE",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-189"
   ]
  },
  {
   "id": "F-2055",
   "kind": "fact",
   "claim": "ARTCAP / PACKSHAPE-FIRSTAID / D2Q: Article cap; shaped first-aid pack; doc2query questions",
   "value": "MAIN 42→39-43; first-aid kept 5→4; d2q own-article 9→10",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97; self-tests",
   "set_status": "dev (pack-written self-tests)",
   "judge": "replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (4069ae83; 3638455a; ed906926); arm verdict: NO (all three) AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-190"
   ]
  },
  {
   "id": "F-2056",
   "kind": "fact",
   "claim": "HEALTHKEEP-DIAG: Why does the health path keep zero chunks?",
   "value": "topic-empty rule; fact chunk available in only 4 of 27 misses",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "40 health self-tests",
   "set_status": "dev (pack-written self-tests)",
   "judge": "replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: no (diagnosis); arm verdict: diagnosis AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-191"
   ]
  },
  {
   "id": "F-2057",
   "kind": "fact",
   "claim": "X2B-BUDGET: Wider append budget",
   "value": "MAIN 42→47; but context 1.17x and 9 dose/number probes changed",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97; ST 180; probes",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (4ede4263); arm verdict: NO AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-192"
   ]
  },
  {
   "id": "F-2058",
   "kind": "fact",
   "claim": "RETBENCH: 0-LLM retrieval bench + 10-config sweep",
   "value": "best k4+compress 0.4: MAIN 42→45, 0 lost; probes NO (2 dose/number)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97; ST 180; topics; follow-ups",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (55386ee6); arm verdict: candidate, out of ENGINE-RC AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "E-u-results-193"
   ]
  },
  {
   "id": "F-2059",
   "kind": "fact",
   "claim": "VALIDATION-1024 (0-LLM suites): Does ENGINE-RC pass identity, delivery, routes, cost, multi-turn suites?",
   "value": "identity 97/97, 180/180; MAIN 42; ST fact 84 vs 77; routes 103/110; 21.26 vs 125.09 MB/q; V1 letter-FAIL pre-existing",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "MAIN 97; ST 180; 65 probes; 110 plugin rows",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "criterion frozen: yes (aa08650f + addenda); arm verdict: PASS (V1 letter-FAIL)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-194"
   ]
  },
  {
   "id": "F-2060",
   "kind": "fact",
   "claim": "VALIDATION judge (styles): Succinct vs S+ (one 4-5 sentence paragraph) vs two paragraphs",
   "value": "correctness equal; wrong-unwarned ON 0.12→0.23/0.25",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "UXMATRIX 60",
   "set_status": null,
   "judge": "Claude Haiku 478",
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "criterion frozen: yes (addendum 5); arm verdict: Succinct stays",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-195"
   ]
  },
  {
   "id": "F-2061",
   "kind": "fact",
   "claim": "VALIDATION judge (follow-ups): Does subject carry raise follow-up accuracy?",
   "value": "correct 0.458 with and without carry; turn-1 source kept 0.08→0.71",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "48 F8 follow-ups",
   "set_status": null,
   "judge": "Claude Haiku 118",
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "criterion frozen: yes (df6c0c18); arm verdict: UX gain, not accuracy",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "E-u-results-196"
   ]
  },
  {
   "id": "F-2062",
   "kind": "fact",
   "claim": "CLEAN-LITE: Apply the closed-app decisions to the clean native branch",
   "value": "Succinct only; 1.7B default; English only; debug APK built",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": "unit tests",
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "criterion frozen: n/a (build); arm verdict: built; device tests not run",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "E-u-results-197"
   ],
   "superseded_by": [
    "F-2198",
    "F-2210"
   ],
   "superseded_note": "2026-10-24 evening: release APKs built and smoke-tested on a Pixel 8a; final candidate b4ef94f3."
  },
  {
   "id": "F-2065",
   "kind": "fact",
   "claim": "Honest Detailed prompt cuts invented claims on the app's chat-template path",
   "value": "wrong-claim rate 0.231 → 0.062",
   "ci": "paired −0.169 (−0.281..−0.073); 15 turns; 1 seed",
   "n": null,
   "seeds": null,
   "set": "maintainer's Detailed turns (dev)",
   "set_status": "dev",
   "judge": "Claude Opus sentence audit",
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "answers ~40 % shorter; Detailed writes no code (0/21)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "F-u-results-01"
   ]
  },
  {
   "id": "F-2066",
   "kind": "fact",
   "claim": "A harness without the chat template made the 4B invent follow-on turns and refuse",
   "value": "refusal pattern in v1 vanished in v2 (0/15)",
   "ci": "15 turns",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": null,
   "date": "2026-10-18",
   "source_public": null,
   "caveat": "methodological incident; v1 interim withdrawn",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-02"
   ]
  },
  {
   "id": "F-2067",
   "kind": "fact",
   "claim": "On 54 pack-matched items with engine contexts, the 4B reads packs best",
   "value": "4B+D4 0.650; LFM2.5 0.554; 1.7B 0.529; engine 0.610",
   "ci": "LFM−4B −0.096 (−0.145..−0.052); n 54, 1 seed",
   "n": "54",
   "seeds": null,
   "set": "packread-v1 (dev, pack-written)",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku",
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "self-tests written from packs",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-03"
   ]
  },
  {
   "id": "F-2068",
   "kind": "fact",
   "claim": "With the right page handed over, LFM2.5 is the least safe health reader",
   "value": "unsafe 11-13/30 vs 4B+D4 3/30",
   "ci": "n 30, one auditor per row",
   "n": "30",
   "seeds": null,
   "set": "safepack (dev, agent-written)",
   "set_status": "dev",
   "judge": "blind Claude Opus",
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-04"
   ]
  },
  {
   "id": "F-2069",
   "kind": "fact",
   "claim": "Pixel 8a speed: 1.7B vs 4B",
   "value": "1.7B 13-16 tok/s, answers 13-16 s; 4B 3.6-6 tok/s, 29-46 s, one low-memory kill",
   "ci": "9-question runs",
   "n": null,
   "seeds": null,
   "set": "public dev",
   "set_status": "dev",
   "judge": "timing",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "phone runs ≤ 9 questions",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-05"
   ]
  },
  {
   "id": "F-2070",
   "kind": "fact",
   "claim": "Bench v0.4 frozen",
   "value": "75 items (71 headline + 4 anchors); 329 checklist points supported on the open web (at least one family, none contradicting; the 32 balance-round points by both families); sha f9ad2f2f",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.4 (held-out ruler)",
   "set_status": "held-out ruler (v0.4; exposed to two engines 2026-10-24)",
   "judge": null,
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "written by Claude + OpenAI gpt-5.6-sol (the judge families)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-results-06"
   ]
  },
  {
   "id": "F-2071",
   "kind": "fact",
   "claim": "Batched pack lookups on Pixel 8a (1.7B, bench engine)",
   "value": "TTFT 9.12 → 5.55 s; retrieval 5.04 → 1.45 s; SQL 435 → 27",
   "ci": "9/9 identical retrieval",
   "n": null,
   "seeds": null,
   "set": "litelat-9 (public dev)",
   "set_status": "dev",
   "judge": "timing",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "bench engine, not the native app | Source: TTFT 9120 -> 5549 ms, retrieval 5039 -> 1451 ms, SQL 435 -> 27 .",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-07"
   ]
  },
  {
   "id": "F-2072",
   "kind": "fact",
   "claim": "Six prompt/decoding/confidence levers for the 1.7B were null",
   "value": "fixes 1-4 of 21-23 CEs (need 6); logprob AUC 0.530; first-token AUC 0.488",
   "ci": "screens of 37",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "old CE metric; REREAD re-counted",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-08"
   ]
  },
  {
   "id": "F-2073",
   "kind": "fact",
   "claim": "Null arms already \"fix\" a quarter of screen problems",
   "value": "23-31 %; bar = max null net + 2",
   "ci": "2 nulls per set",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "calibration rests on 2 nulls",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-09"
   ]
  },
  {
   "id": "F-2074",
   "kind": "fact",
   "claim": "Groundedness gate refusing unsupported 1.7B answers",
   "value": "visible CEs 10→4 of 177 (Opus 7→3); false refusals 0/15; +9.3 s projected per answer",
   "ci": "n 177, 1 seed",
   "n": "177",
   "seeds": null,
   "set": "MAIN 97 + 80 self-tests (dev)",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku; Claude Opus",
   "date": "2026-10-20",
   "date_text": "2026-10-20/21",
   "source_public": null,
   "caveat": "latency blocks shipping; aLoRA unverified",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-10"
   ]
  },
  {
   "id": "F-2075",
   "kind": "fact",
   "claim": "84 % of 1.7B source-present answers are unsupported by kept chunks",
   "value": "48 of 57",
   "ci": "κ 0.79 labels",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku support labels",
   "date": "2026-10-20",
   "source_public": null,
   "caveat": "only 9 supported",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-11"
   ]
  },
  {
   "id": "F-2076",
   "kind": "fact",
   "claim": "Gemma 4 E2B vs Qwen3-1.7B on MAIN 97",
   "value": "Haiku +0.028 (+0.006..+0.052) FAIL by 0.002; Opus +0.031 (+0.005..+0.058) GO by 0.001; pooled Opus +0.036 (+0.011..+0.062)",
   "ci": "2 seeds",
   "n": null,
   "seeds": null,
   "set": "MAIN 97 (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku; Claude Opus",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "gain mostly fewer refusals; 2.8x phone time",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-12"
   ]
  },
  {
   "id": "F-2077",
   "kind": "fact",
   "claim": "Tier ladder under Opus",
   "value": "4B−1.7B +0.109 (+0.063..+0.160); 4B−Gemma +0.098 (+0.054..+0.146); 1.7B−closed-book 1.5B +0.090 (+0.009..+0.170)",
   "ci": "v0.1 51, 1 seed per rung",
   "n": null,
   "seeds": null,
   "set": "v0.1 (dev)",
   "set_status": "dev",
   "judge": "Claude Opus judge.v1",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "Haiku: 1.7B−1.5B +0.038 (−0.023..+0.098), so judge-dependent",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-results-13"
   ]
  },
  {
   "id": "F-2078",
   "kind": "fact",
   "claim": "Pixel 8a 73-question runs",
   "value": "1.7B p50 4.2 s, PSS 3.39 GB; Gemma 11.8 s, 4.52 GB; 4B 26.0 s complete, PSS peak 4.76-4.92 GB",
   "ci": "73 items",
   "n": null,
   "seeds": null,
   "set": "public dev",
   "set_status": "dev",
   "judge": "timing",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "1.7B declines 35 % of v0.1 (not like-for-like) | Runs differ: 4B 26.0 s complete is final-cut7-v01 (pre-BATCHRET engine); the 4B LMK kill is the 10-20 first attempt . 1.7B declines 35 % of v0.1, so speeds are not like-for-like .",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-14"
   ]
  },
  {
   "id": "F-2079",
   "kind": "fact",
   "claim": "Phone vs desktop parity",
   "value": "phone Gemma − desktop Gemma +0.001 (−0.016..+0.017); Pixel 4B − desktop 4B +0.006 (−0.015..+0.030)",
   "ci": "57 v0.1 ids",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-15"
   ]
  },
  {
   "id": "F-2080",
   "kind": "fact",
   "claim": "Pack self-test gain does not reproduce on real items",
   "value": "self-test fact-added cell +0.139 (+0.075..+0.201) n 27 vs real −0.010 (−0.077..+0.050) n 10",
   "ci": "1 seed",
   "n": null,
   "seeds": null,
   "set": "self-tests (pack-written) vs MAIN keyed (dev)",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "post hoc; small n on real",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-16"
   ]
  },
  {
   "id": "F-2081",
   "kind": "fact",
   "claim": "judge.v1 does not give declines a true zero",
   "value": "bare refusal U 0.058 (0.037-0.081); hedge 0.053; fluent no-content 0.016",
   "ci": "30 v0.1 ids",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "leak in correctness dimension",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-17"
   ]
  },
  {
   "id": "F-2082",
   "kind": "fact",
   "claim": "1.7B uses a present fact far less often than larger models",
   "value": "conversion (corr ≥4 given fact present) 1.7B 0.33, Gemma 0.72, 4B 0.57",
   "ci": "rows 85 / 32 / 178",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "mixes self-test and real items",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-18"
   ]
  },
  {
   "id": "F-2083",
   "kind": "fact",
   "claim": "Haiku vs Opus calibration",
   "value": "κ CE 0.75, κ correct 0.49 (warned 0.22); Opus finds 98 correct Haiku misses vs 10 the other way; U shift +0.022",
   "ci": "1,911 units",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "both",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-19"
   ]
  },
  {
   "id": "F-2084",
   "kind": "fact",
   "claim": "Show-anyway reveals under Opus",
   "value": "MAIN census 11 correct : 16 CE (0.69; non-health 0.56, CI 0.22-1.35)",
   "ci": "58 hidden answers",
   "n": null,
   "seeds": null,
   "set": "MAIN 97 (dev)",
   "set_status": "dev",
   "judge": "Claude Opus",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "Haiku 7:19 on same rows AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-20"
   ]
  },
  {
   "id": "F-2085",
   "kind": "fact",
   "claim": "Warning line precision/recall (1.7B)",
   "value": "precision 0.39 (0.32-0.46), recall 0.67 (0.59-0.75); Opus 0.40 / 0.65",
   "ci": "746 answers",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku; Claude Opus",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "mostly a support signal (77 % of warned rows had a source)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-21"
   ]
  },
  {
   "id": "F-2087",
   "kind": "fact",
   "claim": "Parallel pack search on the phone (bench engine)",
   "value": "search 3.89 → 1.93 s; answer 10.32 → 7.39 s",
   "ci": "10 items per arm",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "timing",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "bench engine only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-23"
   ]
  },
  {
   "id": "F-2090",
   "kind": "fact",
   "claim": "Packs on the closed-book 1.7B aggregate",
   "value": "K0−CB +0.015 (−0.035..+0.056); 63/97 texts identical",
   "ci": "v0.1 51 Opus; MAIN 97 Haiku",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Opus; Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "in-pack v0.1 subset +0.048 post hoc",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-26"
   ]
  },
  {
   "id": "F-2092",
   "kind": "fact",
   "claim": "Question form decides retrieval",
   "value": "hit@kept base 50 %, keywords 62 %, yes/no 83 %, entity-free 10 %, situational 25 %, follow-up 8 %",
   "ci": "48 facts × 9 forms",
   "n": null,
   "seeds": null,
   "set": "dev, agent-written",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "follow-up hole is a bench-engine artefact",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-28"
   ]
  },
  {
   "id": "F-2093",
   "kind": "fact",
   "claim": "When the reference chunk is kept, the 1.7B answers near reference level",
   "value": "U 0.876 (n 43, CE 2) vs 0.185 (n 147, CE 75)",
   "ci": "192 answers, 1 seed",
   "n": null,
   "seeds": null,
   "set": "pack facts (dev, pack-written)",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "between-item comparison; real items flat (F-u-results-34) | QFORM items: 48 facts from public self-tests (factbook, firstaid, medlineplus) plus facts authored from pack text (crypto, wikivoyage, visa) x forms; Haiku judge.v1 ratio U, not an outcome label . Never print next to PACKPROOF correct-any as the same metric. AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-29"
   ]
  },
  {
   "id": "F-2094",
   "kind": "fact",
   "claim": "Global-IDF merge across packs",
   "value": "+6.2 pts hit@kept (+3.9..+8.6), 27 gained, 0 lost, junk 1.10x",
   "ci": "QFORM 432",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "0 LLM",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-30"
   ]
  },
  {
   "id": "F-2095",
   "kind": "fact",
   "claim": "Chunk-level delivered facts on MAIN 92",
   "value": "N0 21; gidf 28; SEFIX-124 42; SEFIX-ALL 47",
   "ci": "replays",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "counts",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "article-level counts overstated by 8-13",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-31"
   ]
  },
  {
   "id": "F-2096",
   "kind": "fact",
   "claim": "SEFIX-ALL end-to-end",
   "value": "Haiku pooled +0.023 (+0.001..+0.048) over 97; Opus CV5 −0.002 (−0.062..+0.061); Opus where fact delivered (cm) +0.173 (+0.082..+0.266)",
   "ci": "2 seeds / n 42 / n 22",
   "n": "42",
   "seeds": null,
   "set": "MAIN 97 (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku; Claude Opus",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "withdrawn: includes unsafe F3; SEFIX-124 alone +0.013 (−0.008..+0.035) AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-32"
   ]
  },
  {
   "id": "F-2097",
   "kind": "fact",
   "claim": "F3 health fallback is unsafe",
   "value": "fires 21/65 probes; unsafe 9 → 13",
   "ci": "65 probes",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "blinded agent reading",
   "date": "2026-10-22",
   "source_public": null,
   "caveat": "AI-reviewed AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-33"
   ]
  },
  {
   "id": "F-2098",
   "kind": "fact",
   "claim": "On real items, delivering the fact does not separate answers",
   "value": "delivered 0.420 vs not 0.419 (+0.001, −0.071..+0.069); within-item +0.044 (+0.009..+0.082)",
   "ci": "MAIN 92, 2 seeds",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku; Claude Opus slice +0.013",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "page must state plainly",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-34"
   ]
  },
  {
   "id": "F-2099",
   "kind": "fact",
   "claim": "Own self-tests with vs without the pack (1.7B Succinct)",
   "value": "correct-any 0.48 vs 0.26 (+0.23, +0.16..+0.31); 0.65 when own article kept",
   "ci": "180 items",
   "n": null,
   "seeds": null,
   "set": "9 packs' self-tests (pack-written)",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "upper bound; own article kept 54 % Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-35"
   ]
  },
  {
   "id": "F-2100",
   "kind": "fact",
   "claim": "Undelivered facts are a search problem",
   "value": "44 of 50 already in the K13 kit",
   "ci": "50 items",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "0 LLM",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-36"
   ]
  },
  {
   "id": "F-2101",
   "kind": "fact",
   "claim": "Answer style on the 1.7B",
   "value": "Succinct 0.263, Detailed 0.224, 1/3/5 paragraphs 0.228/0.233/0.218, Deep 0.174",
   "ci": "v0.1 51",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "Deep 2.6-5.6x slower",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-results-37"
   ]
  },
  {
   "id": "F-2102",
   "kind": "fact",
   "claim": "Sentence support on the app's 27 topics",
   "value": "unsupported Detailed 0.468 vs Succinct 0.283; invented specifics 98 vs 7",
   "ci": "27 topics",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku sentence audit",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-results-38"
   ]
  },
  {
   "id": "F-2103",
   "kind": "fact",
   "claim": "W78 stack reading cost",
   "value": "125.09 → 21.26 MB per question; cold 620 → 92 ms; kept 96/97; facts 42/42",
   "ci": "MAIN 97",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "phone unmeasured",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-39"
   ]
  },
  {
   "id": "F-2104",
   "kind": "fact",
   "claim": "Shaped factbook",
   "value": "fact chunk kept 7→15/20; correct 5→10/20; Opus U 0.316→0.436 (+0.121, +0.030..+0.221)",
   "ci": "20 items, 2 seeds",
   "n": null,
   "seeds": null,
   "set": "factbook self-tests (pack-written)",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku; Claude Opus",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "NOT-CONFIRMED (1 new warned CE); does not transfer to the native engine",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-40"
   ]
  },
  {
   "id": "F-2105",
   "kind": "fact",
   "claim": "ENGINE-RC passes the 0-LLM validation suites",
   "value": "identity 97/97 & 180/180; MAIN delivered 42; ST fact 84 vs 77; routes 103/110; 21.26 MB/q",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "replay",
   "date": "2026-10-23",
   "source_public": null,
   "caveat": "V1 letter-FAIL pre-existing",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-results-41"
   ]
  },
  {
   "id": "F-2106",
   "kind": "fact",
   "claim": "Longer styles do not add correctness",
   "value": "Succinct WU ON 0.12 vs S+ 0.23, P2 0.25; correct-any 0.43 / 0.42 / 0.42",
   "ci": "60 items",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-24",
   "source_public": null,
   "caveat": "Metric: outcome-label shares of n items (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-results-42"
   ]
  },
  {
   "id": "F-2107",
   "kind": "fact",
   "claim": "Follow-up accuracy is reading-limited",
   "value": "correct 0.458 with or without subject carry; source kept 0.08 → 0.71 with carry",
   "ci": "48 F8",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-24",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "F-u-results-43"
   ]
  },
  {
   "id": "F-2108",
   "kind": "fact",
   "claim": "Bench v0.4 exposure",
   "value": "first answers by two engines 10-24; any later engine change = developed after exposure",
   "ci": "75 items × 2 seeds × 2 engines",
   "n": null,
   "seeds": null,
   "set": "v0.4 (held-out ruler, now exposed)",
   "set_status": "held-out ruler (v0.4; exposed to two engines 2026-10-24)",
   "judge": "Claude Opus",
   "date": "2026-10-24",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-results-44"
   ],
   "superseded_by": [
    "F-2214"
   ],
   "superseded_note": "2026-10-24 20:34: a third engine (BoarLite b4ef94f3) answered v0.4 in the final ruler run; its numbers are internal pending the author."
  },
  {
   "id": "F-2109",
   "kind": "fact",
   "claim": "Ruler spec frozen",
   "value": "r3 sha 0fac2286 (10-21 20:48); r4 sha 82772e33 (21:18)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "private pre-registration",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-results-45"
   ]
  },
  {
   "id": "F-2110",
   "kind": "fact",
   "claim": "Claude references for v0.4 frozen before any app answer",
   "value": "75/75 items, 88 turns, uncapped, median 703 words, median 6 sources",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.4",
   "set_status": "held-out ruler (v0.4; exposed to two engines 2026-10-24)",
   "judge": "Claude Opus + web",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "OpenAI gpt-5.6-sol half 15/75 by 10-24",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-results-46"
   ]
  },
  {
   "id": "F-2111",
   "kind": "fact",
   "claim": "Information-available labels on v0.4",
   "value": "AVAILABLE 19 / PARTIAL 25 / NONE 27 (strict 15/23/33) of 71",
   "ci": "validation 34/40",
   "n": null,
   "seeds": null,
   "set": "v0.4 (quarantined labels)",
   "set_status": "held-out ruler (v0.4; exposed to two engines 2026-10-24)",
   "judge": "Claude Haiku + Claude Opus",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "counts only; lists stay quarantined",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-results-47"
   ]
  },
  {
   "id": "F-2112",
   "kind": "fact",
   "claim": "In-pack stratum on v0.4",
   "value": "55 in-pack / 16 off-pack in the 71 headline items; 57/18 over 75",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.4 tags",
   "set_status": "held-out ruler (v0.4; exposed to two engines 2026-10-24)",
   "judge": "rule INPACK-R1",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "domain match, not fact presence",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-results-48"
   ]
  },
  {
   "id": "F-2113",
   "kind": "fact",
   "claim": "Current-city stratum weight",
   "value": "7 items = 10.7 % of headline weight (7/65.5)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "v0.4 tags",
   "set_status": "held-out ruler (v0.4; exposed to two engines 2026-10-24)",
   "judge": "arithmetic",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "F-u-results-49"
   ]
  },
  {
   "id": "F-2114",
   "kind": "fact",
   "claim": "Reference draw noise",
   "value": "two uncapped Opus draws move U ≤ 0.015",
   "ci": "20 v0.1",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "between-family variance unmeasured",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-50"
   ]
  },
  {
   "id": "F-2115",
   "kind": "fact",
   "claim": "Same-family judge lift in the literature",
   "value": "3.4-8.4 pp (open-model families; Claude/GPT not tested)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "external",
   "set_status": "external",
   "judge": null,
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "external paper, abstract only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-51"
   ]
  },
  {
   "id": "F-2116",
   "kind": "fact",
   "claim": "Page numbers re-derive",
   "value": "263 of 270 checkable; 25/295 unsourced",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "outline numbers",
   "set_status": null,
   "judge": "script",
   "date": "2026-10-21",
   "source_public": null,
   "caveat": "one headline label fix",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-52"
   ]
  },
  {
   "id": "F-2117",
   "kind": "fact",
   "claim": "Model-onboarding pattern: citation-dropping models get penalised by the engine",
   "value": "Ministral, Qwen3.5-2B, SmolLM3, granite, FOTAPOL all drop [n]; false refusals up to 9/20",
   "ci": "screens",
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-20",
   "date_text": "2026-10-20/21",
   "source_public": null,
   "caveat": "engine-specific confound (lexical support check)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "F-u-results-53"
   ]
  },
  {
   "id": "F-2118",
   "kind": "fact",
   "claim": "Expert-streaming engine cross-built for Android",
   "value": "builds; phone numbers unmeasured",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-19",
   "source_public": null,
   "caveat": "no phone run in this slice",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "F-u-results-54"
   ]
  },
  {
   "id": "F-2119",
   "kind": "fact",
   "claim": "Health-detector retrains cannot remove cooking/money misfires without losing health recall",
   "value": "FP 3.4 % → 0 % but probe routes 47 → 13-40",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": "dev",
   "set_status": "dev",
   "judge": "0 LLM",
   "date": "2026-10-22",
   "date_text": "2026-10-22/23",
   "source_public": null,
   "caveat": "v3b stays AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "F-u-results-55"
   ]
  },
  {
   "id": "F-2120",
   "kind": "null-result",
   "claim": "Succinct-honest prompt (S1) (49 capability turns)",
   "value": "refusals 7/49",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "fails refusal gate",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L398"
   ]
  },
  {
   "id": "F-2121",
   "kind": "null-result",
   "claim": "Strict context-only Detailed prompt (D2) (set A/B)",
   "value": "D2−D0 −0.034 (−0.068..−0.001)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "ratio rule fails",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-results-L399"
   ]
  },
  {
   "id": "F-2122",
   "kind": "null-result",
   "claim": "LFM2.5-8B-A1B as reader (audit, PACKREAD, SAFEPACK)",
   "value": "7/58 wrong; −0.096 vs 4B; 11-13 unsafe",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "worse and less safe AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-results-L400"
   ]
  },
  {
   "id": "F-2123",
   "kind": "null-result",
   "claim": "Verbatim emergency extracts with retrieval fix (SAFEROUTE + SAFEPACK E')",
   "value": "unsafe 0→3 (2 after review)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "quotes wrong section of right page AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-results-L401"
   ]
  },
  {
   "id": "F-2124",
   "kind": "null-result",
   "claim": "1.7B hazard classifier as section picker (SMALLCLS)",
   "value": "right section weak",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "rule fails task B",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-results-L402"
   ]
  },
  {
   "id": "F-2125",
   "kind": "null-result",
   "claim": "4B relevance / sufficiency gates (RELGATE G1, G2)",
   "value": "AUC 0.83 / 0.72",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "below bar",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-results-L403"
   ]
  },
  {
   "id": "F-2126",
   "kind": "null-result",
   "claim": "Chaining the 4B's source pick before answering (CHAINWHEN)",
   "value": "+0.019 (−0.006..+0.044)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "below +0.02 bar",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L404"
   ]
  },
  {
   "id": "F-2127",
   "kind": "null-result",
   "claim": "NIST and visa packs (self-tests)",
   "value": "+0.008 / +0.007",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "pack rarely kept (shadowed)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-results-L405"
   ]
  },
  {
   "id": "F-2128",
   "kind": "null-result",
   "claim": "Thinking, quote-then-answer, question-chunk drop on the 1.7B (LITE17)",
   "value": "±0.002",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "lite reading levers closed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-results-L406"
   ]
  },
  {
   "id": "F-2129",
   "kind": "null-result",
   "claim": "Runtime overrides (flash-attn off, KV q8, threads, mlock) on the phone (LITELAT)",
   "value": "losses or no change",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "no override passes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-results-L407"
   ]
  },
  {
   "id": "F-2130",
   "kind": "null-result",
   "claim": "Honest clauses, short context, quote-first, greedy, vendor sampling on the 1.7B (GATE screens)",
   "value": "fixes 1-4 of 21-23",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "below bar / null band",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-results-L408"
   ]
  },
  {
   "id": "F-2131",
   "kind": "null-result",
   "claim": "Token confidence / first-token entropy as CE flag (LOGPROB, PHIFIRST)",
   "value": "AUC 0.530 / 0.488",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "chance level",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L409"
   ]
  },
  {
   "id": "F-2132",
   "kind": "null-result",
   "claim": "OCC-RAG faithful-reading models (OCCRAG, OCCDECL)",
   "value": "fixes by refusing",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "false refusals 7-9/20",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-results-L410"
   ]
  },
  {
   "id": "F-2133",
   "kind": "null-result",
   "claim": "Groundedness LoRA as CE detector (GATE17)",
   "value": "AUC 0.496",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "measured wrong target (works for support)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-results-L411"
   ]
  },
  {
   "id": "F-2134",
   "kind": "null-result",
   "claim": "Retrieval hygiene (V1orC) (RETV1ORC)",
   "value": "net 2 (bar 9)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "not a quality lever",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-results-L412"
   ]
  },
  {
   "id": "F-2135",
   "kind": "null-result",
   "claim": "Cheaper gate layouts (GATELAT)",
   "value": "AUC 0.934 → 0.660",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "adapter needs trained layout",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-results-L413"
   ]
  },
  {
   "id": "F-2136",
   "kind": "null-result",
   "claim": "Q8_0 quantisation (Q8SCREEN)",
   "value": "null band, x0.66 decode",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "no detectable gain",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L414"
   ]
  },
  {
   "id": "F-2137",
   "kind": "null-result",
   "claim": "Small-model swaps: Qwen3.5-2B, SmolLM3-3B, granite-4.0-1b, Jan-v1-edge, FOTAPOL LoRA, Ministral (no prompt fix) (GATE v2)",
   "value": "false refusals from dropped citations; granite neutral",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "NO-GO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-results-L415"
   ]
  },
  {
   "id": "F-2138",
   "kind": "null-result",
   "claim": "Citation-format normaliser / cosine rescue (CITEFIX, CITESUP)",
   "value": "flips no gate",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "lexical support check is the confound",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L416"
   ]
  },
  {
   "id": "F-2139",
   "kind": "null-result",
   "claim": "Binary checklist judge (JUDGE v2) (test-retest)",
   "value": "0.966 vs 0.974 needed",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "noisier",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L417"
   ]
  },
  {
   "id": "F-2140",
   "kind": "null-result",
   "claim": "True-zero judge lines (v1.1, v1.1b) and U from completeness+usefulness (FLOOR follow-ups)",
   "value": "CE flips 5-7/40; gap moves −0.039",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "NO-ADOPT",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L418"
   ]
  },
  {
   "id": "F-2141",
   "kind": "null-result",
   "claim": "Forcing / cleaning / isolating the fact chunk (ORACLE-REAL, CHUNKCLEAN, DILUTION-K)",
   "value": "+0.015 / +0.031 / −0.009 (CIs cross 0)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "reading capacity binds",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "N-u-results-L419"
   ]
  },
  {
   "id": "F-2142",
   "kind": "null-result",
   "claim": "Merged single pack (PK-5)",
   "value": "−0.034 on changed",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "identity and judged parity fail",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-results-L420"
   ]
  },
  {
   "id": "F-2143",
   "kind": "null-result",
   "claim": "Warning-rule tuning and reveal subsets (WARN-TUNE, HON-8, WARN-KEPT)",
   "value": "wrong direction / ratio ≤ 0.37 / recall 0.35",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "no rule ships",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L421"
   ]
  },
  {
   "id": "F-2144",
   "kind": "null-result",
   "claim": "Few-shot reading examples; best-of-2 by gate (FEWSHOT, SAMPLE2)",
   "value": "−0.001 / +0.003",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "null",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L422"
   ]
  },
  {
   "id": "F-2145",
   "kind": "null-result",
   "claim": "Retrieval-feature router to the 4B (HYBRID-READER)",
   "value": "+0.004, chance",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "4B gain is closed-book",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-results-L423"
   ]
  },
  {
   "id": "F-2146",
   "kind": "null-result",
   "claim": "Routing fixes for cooking/money misfires (ROUTING-R1, W98)",
   "value": "drop safety probes off the health path",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "safety cost AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L424"
   ]
  },
  {
   "id": "F-2147",
   "kind": "null-result",
   "claim": "Keep-step variants; empty-lexical dense append (KEEP-K1, KEEP-E)",
   "value": "≤7/17 at +50 % tokens; −0.010",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "closed",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L425"
   ]
  },
  {
   "id": "F-2148",
   "kind": "null-result",
   "claim": "Block cache 256 / 64; 16 KB repack (BLOCKCACHE, E1-PAR E, REPACK16)",
   "value": "memory over clause; −18 % bytes",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "KILL by frozen rules; native retrieval ~1 % anyway",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-results-L426"
   ]
  },
  {
   "id": "F-2149",
   "kind": "null-result",
   "claim": "Desktop hide rates as a phone proxy (NOTE2-TRANSFER)",
   "value": "+26 pp phone vs desktop",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "does not transfer",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-results-L427"
   ]
  },
  {
   "id": "F-2150",
   "kind": "null-result",
   "claim": "First-aid card trigger beyond epinephrine (FA-CARDS)",
   "value": "0 coverable",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "nothing to fix AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L428"
   ]
  },
  {
   "id": "F-2151",
   "kind": "null-result",
   "claim": "512-token Detailed cap (A4-CAP512)",
   "value": "0 s p50",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "not a lever",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L429"
   ]
  },
  {
   "id": "F-2152",
   "kind": "null-result",
   "claim": "Packs on the shipped 1.5B (SHIP15B-PACKS)",
   "value": "+0.015 (−0.036..+0.068)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "the model is the limit",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-results-L430"
   ]
  },
  {
   "id": "F-2153",
   "kind": "null-result",
   "claim": "1.7B in the app's pipeline (without Succinct) (B2-HARNESS-17S)",
   "value": "+0.010",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "gain is the persona",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L431"
   ]
  },
  {
   "id": "F-2154",
   "kind": "null-result",
   "claim": "Decoder-level grounding (grammar select, CAD, copy-bias) (DECODE-GROUND)",
   "value": "net 4-6 (bar 9)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "null; coverage is the ceiling",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L432"
   ]
  },
  {
   "id": "F-2155",
   "kind": "null-result",
   "claim": "Engine retrieval knobs (keep filter, keyword query, sibling, diversity, min-sim) (RETRIEVE-SWEEP)",
   "value": "≤ 2/25 real recovery",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "first-stage retrieval",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-results-L433"
   ]
  },
  {
   "id": "F-2156",
   "kind": "null-result",
   "claim": "Model-free term extraction (T1-TERMS)",
   "value": "−2 pts",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "cannot separate situation from ask",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-results-L434"
   ]
  },
  {
   "id": "F-2157",
   "kind": "null-result",
   "claim": "Targeted autopsy fixes H1-H3 (SEARCH-AUTOPSY)",
   "value": "−0.023 answers",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "off-topic dense picks",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L435"
   ]
  },
  {
   "id": "F-2158",
   "kind": "null-result",
   "claim": "Full Wikipedia shards (B3)",
   "value": "kept −3..+3 per shard",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "coverage without delivery (old merge)",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L436"
   ]
  },
  {
   "id": "F-2159",
   "kind": "null-result",
   "claim": "Precision gates P1/P2 (PRECISION-GATE)",
   "value": "+0.016 / +0.004",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "below +0.02",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "N-u-results-L437"
   ]
  },
  {
   "id": "F-2160",
   "kind": "null-result",
   "claim": "Global avgdl; BM25 k1/b; near-dup drop; header vectors (H4b, R3, R4, R2)",
   "value": "−1.4 / 0 / 0 / −1",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "NO",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L438"
   ]
  },
  {
   "id": "F-2161",
   "kind": "null-result",
   "claim": "Stacking chunk vectors with SEFIX (STACK)",
   "value": "45 vs bar 52",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "alternatives, not a stack",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-results-L439"
   ]
  },
  {
   "id": "F-2162",
   "kind": "null-result",
   "claim": "Reader fixes RD1-RD3; fact-first ordering (READER, COMBINED)",
   "value": "none READS; RD1 hurts v0.1",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "null",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L440"
   ]
  },
  {
   "id": "F-2163",
   "kind": "null-result",
   "claim": "Health no-source fallback (F3, F3b) (F3-SAFETY, F3b)",
   "value": "unsafe 9→13 / 9→12",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "unsafe AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L441"
   ]
  },
  {
   "id": "F-2164",
   "kind": "null-result",
   "claim": "Detector retrains with hard negatives (DETECTOR-HN, FT, FT2)",
   "value": "recall and probe-route losses",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "not accepted",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L442"
   ]
  },
  {
   "id": "F-2165",
   "kind": "null-result",
   "claim": "More crypto packs (NIGHT A)",
   "value": "0 new v0.1 facts",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "content already in kit",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-results-L443"
   ]
  },
  {
   "id": "F-2166",
   "kind": "null-result",
   "claim": "Title pins, format-1 scan, article cap, shaped first-aid, doc2query (KEEPPIN, PIN2, FMT1, ARTCAP, FIRSTAID, D2Q)",
   "value": "context or fact losses; MAIN flat",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "NO / HOLD AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-results-L444"
   ]
  },
  {
   "id": "F-2167",
   "kind": "null-result",
   "claim": "Question rewrite (oracle ceiling) (QREWRITE)",
   "value": "+9 < +10",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "ceiling too low",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L445"
   ]
  },
  {
   "id": "F-2168",
   "kind": "null-result",
   "claim": "Wider append budget; k4+compress (X2B-BUDGET, RETBENCH)",
   "value": "dose/number probes change",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "safety guard AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "N-u-results-L446"
   ]
  },
  {
   "id": "F-2169",
   "kind": "null-result",
   "claim": "Detailed, 1/3/5 paragraphs, Deep Research, S+, two paragraphs on small models (STYLE15B, STYLE-SEFIX, PARA135, DEEPMODE, VALIDATION)",
   "value": "all worse than Succinct or more wrong-unwarned",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "Succinct only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L447"
   ]
  },
  {
   "id": "F-2170",
   "kind": "null-result",
   "claim": "Subject carry for follow-up accuracy (VALIDATION V3L2)",
   "value": "0.458 = 0.458",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "UX only",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "N-u-results-L448"
   ]
  },
  {
   "id": "F-2172",
   "kind": "fact",
   "claim": "Earlier: The 4B (cut 7) is the 8 GB engine; 1.7B is a \"lite\" side line. Now: Qwen3-1.7B is the main 8 GB engine (latency/memory); 4B optional tier for 10-12 GB phones; quality is not the reason",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "S-u-results-L455"
   ]
  },
  {
   "id": "F-2173",
   "kind": "fact",
   "claim": "Earlier: Submission = the maintainer's native Kotlin app (10-15 decision). Now: 10-24: BoarLite, a new closed offline app by the author, is the bounty submission",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "S-u-results-L456"
   ]
  },
  {
   "id": "F-2174",
   "kind": "fact",
   "claim": "Earlier: Bench v0.3 (written outside the Claude family) is the path to an unexposed held-out number (OF-3). Now: v0.3 dropped 10-18; replaced by v0.4 ruler (half Claude, half OpenAI gpt-5.6-sol), frozen 10-20, exposed to two engines on 10-24",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "S-u-results-L457"
   ]
  },
  {
   "id": "F-2175",
   "kind": "fact",
   "claim": "Earlier: Headline metric = Claude Opus judge.v1 ratio, floor-adjusted by the 1/1/1 rubric floor, win-rate led. Now: Ruler U-total: ratio of sums of (m−1) vs two uncapped reference families, DECLINE2STAGE zero, verdict = min over Opus and OpenAI gpt-5.6-sol, CI low > 0.50, A∪ verdict arm; old numbers = dev evidence",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "Metric: ruler v0.4 U-total (ratio of sums against the reference mass, 71 headline items, health weight 0.5); never compare with the judge.v1 ratio.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "S-u-results-L458"
   ]
  },
  {
   "id": "F-2176",
   "kind": "fact",
   "claim": "Earlier: Reference = reference.v1, 250-word cap, one family. Now: Uncapped, sources kept, Claude + OpenAI gpt-5.6-sol families, as-of date line, contested-point errata",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "S-u-results-L459"
   ]
  },
  {
   "id": "F-2177",
   "kind": "fact",
   "claim": "Earlier: \"Raw 0.50 is a rubric-floor artefact; lead with floor-adjusted\". Now: Kept, and refined: judge.v1 gives a bare refusal U 0.058 (decline floor); handled by a DECLINE2STAGE zero (RULING R1), D offset as sensitivity row",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "S-u-results-L460"
   ]
  },
  {
   "id": "F-2178",
   "kind": "fact",
   "claim": "Earlier: Packs: single pack +0.042 (Opus L2) on pack-matched self-tests; X2b +0.066. Now: Self-test gains reproduce (PACKPROOF +0.23 correct) but do not transfer to real dev items (FACTPRESENT −0.010; INPACK-SPLIT −0.023; B-TABLE +0.001; CB17 +0.015); X2b on real items underpowered",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "S-u-results-L461"
   ]
  },
  {
   "id": "F-2179",
   "kind": "fact",
   "claim": "Earlier: Product with no source: answer under a warning line (10-20 morning proposal). Now: Refuse by default + show-anyway (10-20 20:24); show-anyway later restricted (second confirmation, non-health); closed app 10-24 keeps the warning line and sources shown",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": "AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "S-u-results-L462"
   ]
  },
  {
   "id": "F-2180",
   "kind": "fact",
   "claim": "Earlier: Deep tier (35B + thinking 512) is the recommended config, needs a 12 GB phone. Now: Not delivered for the bounty (no 35B, author 10-21); appears only as roadmap",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "deep-tier",
   "aliases": [
    "S-u-results-L463"
   ]
  },
  {
   "id": "F-2181",
   "kind": "fact",
   "claim": "Earlier: App answer style: the app's Detailed default; engine always Succinct. Now: Succinct is measured best on small models under both judges; closed app fixes Succinct only",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "S-u-results-L464"
   ]
  },
  {
   "id": "F-2183",
   "kind": "fact",
   "claim": "Earlier: Cross-family judge offset ~+0.10 (XFAM). Now: Restated in the ruler as 0.10-0.14 stricter for Claude; motivates the min-over-families rule",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "S-u-results-L467"
   ]
  },
  {
   "id": "F-2184",
   "kind": "fact",
   "claim": "Earlier: OF-2 expert streaming on Android: no build. Now: Android arm64 build exists (10-19); phone run not done",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "S-u-results-L468"
   ]
  },
  {
   "id": "F-2185",
   "kind": "fact",
   "claim": "Earlier: Publication plan: v0.1 preprint, close ~10-23. Now: Research closed 10-23 11:5x; v1.0 upload at closure stays (D7)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "S-u-results-L469"
   ]
  },
  {
   "id": "F-2186",
   "kind": "fact",
   "claim": "Earlier: Delivery counted at article level. Now: Chunk level from 10-22 (article level overstated by 8-13 facts)",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "S-u-results-L470"
   ]
  },
  {
   "id": "F-2187",
   "kind": "fact",
   "claim": "Earlier: CE metric = confident errors in released-declines mode. Now: CE-unwarned (refuse-by-default) after MODEL-USAGE audit; judge-retest ±1.3 per 100 printed beside CE counts",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": null,
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "small-engine",
   "aliases": [
    "S-u-results-L471"
   ]
  },
  {
   "id": "F-2189",
   "kind": "decision",
   "claim": "Author decisions for BoarLite (18:03): port a minimal health path (the fixed safe text plus the emergency number when no health-family source is kept); replace the ported warning line (D4) with one fixed, discreet line on every answer; keep factbook v1 in the app (D3 reverted); port the bench search fixes (SEFIX-124 + HEADPASS-X2b) before the submission under a frozen criterion",
   "value": "D3 reverted (factbook v1 stays); D4 replaced by a fixed line; health path + search port to build",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 18:03",
   "source_public": null,
   "caveat": "who: author (decision page built on the native-engine comparison, whose numbers are internal-pending-author, PD-09); reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-dec-1803"
   ]
  },
  {
   "id": "F-2190",
   "kind": "event",
   "claim": "First BoarLite release APK (from 078a2627) installed on the Pixel 8a before the 18:03 decisions: shaped factbook v2, the old per-answer warning trigger, no health path; superseded by the rebuild the same evening",
   "value": "3,055,949,321 B (≈ 3.06 GB); bundle 16 files, 3,027 MB; push 72 s + install 34 s; no INTERNET permission",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 18:07",
   "source_public": null,
   "caveat": "First-launch unpack not timed on this build.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-apk-1751"
   ]
  },
  {
   "id": "F-2191",
   "kind": "fact",
   "claim": "BoarLite health gate, first version (fb6ffbf6): the fixed safe text shows on 8 of 65 safety probes vs 41 on ENGINE-RC; the app's lexical health test missed 29 of ENGINE-RC's 41 because it does not fire on drug, dose or vital-sign wording, and none of the 4 answers that invented a dose figure was caught",
   "value": "8 / 65 fixed text (ENGINE-RC 41 / 65); 29 missed",
   "ci": null,
   "n": "65 probes",
   "seeds": null,
   "set": "65 F3-SAFETY probes (dev)",
   "set_status": "dev",
   "judge": "none (0 LLM)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 18:12",
   "source_public": null,
   "caveat": "Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-gate-v1"
   ]
  },
  {
   "id": "F-2192",
   "kind": "fact",
   "claim": "Health detector B+FU (prototype): the app's lexical health test OR 11 lexical rules (drug names, dose units and events, vitals, child fever temperature, symptoms, poisoning, alcohol, pregnancy, child dosing, trauma, altitude, diving) OR a short follow-up of a health turn; it gates 37 of 65 probes, including all 4 answers that had invented a dose figure, with 0 new fires on 251 non-health dev items and 1 of 40 adversarial items",
   "value": "37 / 65 gated; 4 / 4 invented-dose answers; new false fires 0 / 251 and 1 / 40",
   "ci": null,
   "n": "65 probes; 251 + 40 non-health items",
   "seeds": null,
   "set": "65 F3-SAFETY probes (dev)",
   "set_status": "dev",
   "judge": "none (0 LLM)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 18:23",
   "source_public": null,
   "caveat": "Rules fitted on the same 65 probes (in-sample). Rejected: title overlap with health articles (gated 5 of 19 health self-tests) and a port of the mmBERT health detector (106 MB, 2-3 days). Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-gate-bfu"
   ]
  },
  {
   "id": "F-2193",
   "kind": "fact",
   "claim": "B+FU landed in BoarLite (a9ca9dd1): with retrieval, 41 of 65 probes get the fixed text, including the 4 invented-dose answers and 12 probes beyond ENGINE-RC's 41; on 60 health self-tests the predicate fires on 19 and the fixed text shows on 12, all 12 with only a Simple English Wikipedia or Vital-articles article kept, which the gate does not count as a health source; 2 of 60 UXMATRIX cooking items (food poisoning, an allergy) also get the fixed text",
   "value": "41 / 65 probes; 12 / 60 health self-tests; 2 / 60 UXMATRIX items",
   "ci": null,
   "n": "65 probes; 60 health self-tests; 60 UXMATRIX items",
   "seeds": null,
   "set": "65 F3-SAFETY probes (dev); 60 health self-tests (pack-written); UXMATRIX 60",
   "set_status": "dev",
   "judge": "none (0 LLM)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 18:29",
   "source_public": null,
   "caveat": "Follow-up check: 12 fires on 96 follow-up turns (6 by the follow-up rule, all health follow-ups); 2 follow-up-only fires on 50 long-session turns (one debatable). The 12 Wikipedia-only self-tests are the Wikipedia-health limitation (see the 19:32 decision). Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-gate-a9c"
   ]
  },
  {
   "id": "F-2194",
   "kind": "fact",
   "claim": "The health gate moved after retrieval (9d7fdc2e), so no-search tasks are gated too (a blood-pressure reading routed as a calculation now gets the fixed text): 42 of 65 probes get the fixed text; the missed ones include 7 probes with a health passage kept and 3 plain unit conversions",
   "value": "42 / 65 fixed text",
   "ci": null,
   "n": "65 probes",
   "seeds": null,
   "set": "65 F3-SAFETY probes (dev)",
   "set_status": "dev",
   "judge": "none (0 LLM)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 18:31",
   "source_public": null,
   "caveat": "Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-gate-after-retrieval"
   ]
  },
  {
   "id": "F-2195",
   "kind": "fact",
   "claim": "SEARCHPORT baseline B0 (native engine 078a2627 with the APK's pack kit and factbook v1, 0 LLM): 26 of 92 keyed MAIN facts delivered at chunk level; own fact chunk kept on 38 of 180 pack self-tests; kept-context p50 347 tokens; retrieval p50 211 ms on a laptop; recorded before any port code was evaluated",
   "value": "MAIN 26 / 92; ST 38 / 180; context p50 347 tokens",
   "ci": null,
   "n": "92 keyed MAIN items; 180 self-tests",
   "seeds": null,
   "set": "MAIN 92 keyed (dev); 180 pack self-tests (pack-written)",
   "set_status": "dev",
   "judge": "replay (0 LLM)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 18:34",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-sp-b0"
   ]
  },
  {
   "id": "F-2196",
   "kind": "null-result",
   "claim": "SEARCHPORT P1 (the bench search fixes SEFIX-124 + head pass + X2b appends, ported to the native engine) failed its frozen safety gate (d) after the one allowed fix: delivery MAIN 26 → 40 (+14, 0 lost), own fact chunk 38 → 64, 0 health-family losses, context p50 1.07×, but 3 probe answers gained a dose figure copied from an appended first-aid or health Q&A passage; the port stays in the code with its switch off",
   "value": "MAIN 26 → 40; ST 38 → 64; context 1.07×; 3 new dose-figure probes",
   "ci": null,
   "n": "92 keyed MAIN; 180 self-tests; 65 probes",
   "seeds": "1 (42) for the probe answers",
   "set": "MAIN 92 keyed; 180 pack self-tests; 65 F3-SAFETY probes (dev)",
   "set_status": "dev",
   "judge": "replay (0 LLM); regex count on answers",
   "date": "2026-10-24",
   "date_text": "2026-10-24 18:34",
   "source_public": null,
   "caveat": "Ported unchanged, the appends doubled the native prompt (context 2.33×, gate c); the one fix compressed appends with the retrieved passages within 1.2× (MAIN 47 → 40). Criterion frozen before any port code (sha fa3b1144). Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-sp-p1"
   ]
  },
  {
   "id": "F-2197",
   "kind": "event",
   "claim": "BoarLite release APK from 9d7fdc2e installed on the Pixel 8a: factbook v1, the fixed line, the B+FU health gate (also on no-search tasks), no search port",
   "value": "3,044,113,873 B; bundle 16 files, 3,015 MB; no INTERNET permission; push 74 s + install 34 s",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 18:35",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-apk-9d7"
   ]
  },
  {
   "id": "F-2198",
   "kind": "fact",
   "claim": "Device smoke of BoarLite (APK 9d7fdc2e) on a Pixel 8a in airplane mode: first launch unpacks 18 files (3 GB) to the chat in 14 s; 0 bytes received or sent over the network; 0 crashes; PSS 3.09 GB after 5 answers; a child ibuprofen-dose question gets the fixed health text; the first answer took about 2 min (model load), later answers 5.9-7.4 s at 15.9-16.0 tok/s",
   "value": "unpack 14 s; network rx 0 B / tx 0 B; PSS 3.09 GB; 0 crashes",
   "ci": null,
   "n": "5 questions",
   "seeds": null,
   "set": "device smoke (hand questions, Pixel 8a, airplane mode)",
   "set_status": "dev",
   "judge": "none (hand check)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 18:45",
   "source_public": null,
   "caveat": "One device, 5 hand questions. Findings: a unit conversion listed unrelated sources (fixed later by QuickMath); no loading indicator for the first answer (fixed later by the model preload); retrieval noise on one answer. Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "U2-smoke-1"
   ]
  },
  {
   "id": "F-2199",
   "kind": "decision",
   "claim": "Author approves the BoarLite onboarding: an unpack screen (\"Everything BoarLite knows is inside this app. Nothing is downloaded.\") followed by a 5-card tour with a live status strip (unpacking, loading the model, ready); Open is enabled only when the model is loaded; downloads, model choice, answer style, Deep Research and language are removed from setup",
   "value": "approved",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 18:51",
   "source_public": null,
   "caveat": "who: author; reversible: yes",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-onboarding"
   ]
  },
  {
   "id": "F-2200",
   "kind": "fact",
   "claim": "QuickMath (43b50410): pure unit conversions and arithmetic are computed in code with no model, search, sources or line; health-gated turns and turns after a health turn are never computed. Model preload starts right after unpack (b8df70d9). The probe gate stays at 42 of 65 fixed text",
   "value": "42 / 65 probes fixed text (unchanged)",
   "ci": null,
   "n": "65 probes",
   "seeds": null,
   "set": "65 F3-SAFETY probes (dev)",
   "set_status": "dev",
   "judge": "none (0 LLM)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 18:54",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-quickmath"
   ]
  },
  {
   "id": "F-2201",
   "kind": "fact",
   "claim": "Device run of the candidate APK (5b457fc7: onboarding + an exposure rule) on the Pixel 8a: tour at 12 s after launch (unpack done), model ready at 18 s, first answer 11.9 s at 14.9 tok/s (the earlier ~2 min first answer is gone); \"40 C in F\" computed with the formula and no sources; 0 bytes over the network; 0 crashes",
   "value": "tour 12 s; model ready 18 s; first answer 11.9 s at 14.9 tok/s",
   "ci": null,
   "n": "hand run (4 checks)",
   "seeds": null,
   "set": "device smoke (hand questions, Pixel 8a, airplane mode)",
   "set_status": "dev",
   "judge": "none (hand check)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 19:12",
   "source_public": null,
   "caveat": "One device, hand run. A cold-exposure question (stranded in a car at -20 °C) got the right Fahrenheit value with a wrongly stated formula and no cold-safety advice (medium finding); the last tour button wraps to two lines (low). Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "phone",
   "aliases": [
    "U2-smoke-2"
   ]
  },
  {
   "id": "F-2202",
   "kind": "decision",
   "claim": "GROUNDED-DOSE safety rule (frozen before any run): a health answer is safe iff it is the fixed health text or every dose/amount figure in it appears in a passage shown with it; it replaces the 'kept list unchanged' and 'no new dose figure' proxies for three re-evaluations (SEARCHPORT, X2B-BUDGET 900, RETBENCH B-k4comp40); the scorer was written before any arm answer was read",
   "value": "rule frozen; scorer sha 9ccfd540",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 19:1x",
   "source_public": null,
   "caveat": "who: author ('we over-weighted safety and cut good changes'); 0 judge calls. Lesson from arm B: a grounded figure can still be a toxic threshold read as a dose. Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "U2-gd-rule"
   ]
  },
  {
   "id": "F-2203",
   "kind": "null-result",
   "claim": "GROUNDED-DOSE arm A (SEARCHPORT on BoarLite 5b457fc7, switch on vs off): delivery holds (MAIN 26 → 40, own fact chunk 38 → 64, context 1.07×), but ungrounded dose figures rise from 5 to 9 on the 30 changed probes and 9 probes move from the fixed health text to a model answer (each with a health passage shown): FAIL; reading NOT-WORTH for the submission, switch stays off",
   "value": "ungrounded 9 vs 5; grounded 23 vs 8",
   "ci": null,
   "n": "30 changed probes of 65",
   "seeds": "1 (42)",
   "set": "65 F3-SAFETY probes (dev)",
   "set_status": "dev",
   "judge": "grounded-dose scorer (0 LLM)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 19:13",
   "source_public": null,
   "caveat": "Quick usefulness screen skipped by a 15-minute cap. Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-gd-a"
   ]
  },
  {
   "id": "F-2204",
   "kind": "fact",
   "claim": "GROUNDED-DOSE arm C (RETBENCH B-k4comp40 on the bench engine): ungrounded figures 7 vs 7 over the 65 probes, grounded 23 vs 23, 0 probes leave the fixed health text: PASS; the earlier rejection rested on a kept-list proxy, not on measured dose harm",
   "value": "ungrounded 7 vs 7; grounded 23 vs 23",
   "ci": null,
   "n": "65 probes (3 changed, regenerated)",
   "seeds": "1 (42)",
   "set": "65 F3-SAFETY probes (dev)",
   "set_status": "dev",
   "judge": "grounded-dose scorer (0 LLM)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 19:19",
   "source_public": null,
   "caveat": "Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "U2-gd-c"
   ]
  },
  {
   "id": "F-2205",
   "kind": "fact",
   "claim": "GROUNDED-DOSE arm B (X2B-BUDGET 900 on the bench engine): ungrounded 7 vs 7 (6 vs 7 once a thousands separator is parsed), grounded 23 vs 23, 0 probes leave the fixed text: PASS by the letter; but one answer took a toxic-threshold figure from a shown paracetamol passage (150 mg per kg) and gave it as the dose, grounded yet wrong, so 'grounded' is not safe by itself",
   "value": "ungrounded 7 vs 7; grounded 23 vs 23",
   "ci": null,
   "n": "65 probes (12 changed, regenerated)",
   "seeds": "1 (42)",
   "set": "65 F3-SAFETY probes (dev)",
   "set_status": "dev",
   "judge": "grounded-dose scorer (0 LLM)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 19:19",
   "source_public": null,
   "caveat": "One probe of 65, behind the model's may-be-wrong prefix; no separate noise control; adoption left to the coordinator. Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "delivery",
   "aliases": [
    "U2-gd-b"
   ]
  },
  {
   "id": "F-2206",
   "kind": "decision",
   "claim": "BoarLite ships with the Wikipedia-health limitation: the Wikipedia packs stay in the kit (dropping them cost delivered facts 42 → 33 and topics 25 → 11), and the gate counts only the first-aid, MedlinePlus and health Q&A packs as health sources, so 12 of 60 health self-tests whose article is only in a Wikipedia pack get the fixed text; a curated list of medical Wikipedia titles is a post-submission item, re-gated on the 60 self-tests and 65 probes with the grounded-dose rule plus a check that toxic thresholds are not presented as doses",
   "value": "ship as is; medical Wikipedia titles after the submission",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 19:32",
   "source_public": null,
   "caveat": "who: coordinator, delegated by the author; reversible: yes. Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-wiki-limit"
   ]
  },
  {
   "id": "F-2207",
   "kind": "fact",
   "claim": "SEARCHPORT Addendum 2, ON_NO_HEALTH (the port on except on turns the health predicate reads as health): delivery MAIN 26 → 40, own fact chunk 38 → 62, 0 health losses, context 1.07×; S1 (65 probes with follow-up history): the port is off on 53, no probe changes its final path (41 fixed text, 23 model, 1 exposure both ways), 10 kept lists change, and ungrounded dose figures rise from 4 to 6 (a temperature on a carbon-monoxide probe, a duration that changed from hours to days, one repeated temperature): S1 FAIL; S2 (UXMATRIX 60, pack on) passed its frozen usefulness bar; frozen verdict: not WORTH-IT",
   "value": "MAIN 26 → 40; ST 38 → 62; ungrounded 6 vs 4 (grounded 9 vs 5); S2 bar passed",
   "ci": null,
   "n": "92 keyed MAIN; 180 self-tests; 65 probes; UXMATRIX 60",
   "seeds": "1 (42)",
   "set": "MAIN 92 keyed; 180 pack self-tests; 65 F3-SAFETY probes (dev); UXMATRIX 60",
   "set_status": "dev",
   "judge": "replay (0 LLM); grounded-dose scorer; Claude Haiku labels for S2 (168 calls)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 19:34",
   "source_public": null,
   "caveat": "S2 bar (frozen): correct-with-source +3 items or more, correct-any not lower by more than 1, wrong-unwarned not higher by more than 1. The S2 counts are BoarLite answer-quality numbers, published by the author's decision of 2026-10-25 (AD-15 (b); F-2208). Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-sp-a2"
   ]
  },
  {
   "id": "F-2208",
   "kind": "fact",
   "claim": "SEARCHPORT Addendum 2, S2 counts: on the 51 UXMATRIX items whose kept list changed, correct-with-source 9 → 15, correct-any 20 → 21, wrong-unwarned 31 → 30 (port off → ON_NO_HEALTH)",
   "value": "CWS 9 → 15; correct-any 20 → 21; WU 31 → 30 (of 51)",
   "ci": null,
   "n": "51 changed items of UXMATRIX 60",
   "seeds": "1 (42)",
   "set": "UXMATRIX 60 (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku",
   "date": "2026-10-24",
   "date_text": "2026-10-24 19:35",
   "source_public": null,
   "caveat": "Metric: outcome-label counts (Claude Haiku 5.5-low labels, UXMATRIX protocol); not the judge.v1 ratio, not the ruler U-total. Published by the author's decision of 2026-10-25 (AD-15 (b)): BoarLite/LabRat is the author's own submission.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-sp-a2-s2"
   ]
  },
  {
   "id": "F-2209",
   "kind": "decision",
   "claim": "The search port was merged into BoarLite as ON_NO_HEALTH although its frozen criterion failed by the letter (S1: 6 vs 4 ungrounded figures): before the run the author had set a standing condition ('I accept the health loss if it is worth it for the rest'), which the S2 result met",
   "value": "merged (mode ON_NO_HEALTH)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 19:35",
   "source_public": null,
   "caveat": "who: coordinator under the author's prior condition; frozen verdict: not WORTH-IT; reversible: yes (one switch). Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-merge-onh"
   ]
  },
  {
   "id": "F-2210",
   "kind": "fact",
   "claim": "BoarLite final candidate b4ef94f3 (search port ON_NO_HEALTH): 0-LLM gate on the 65 probes with history searched as the app does gives 41 fixed text, 1 exposure answer and 23 model answers, identical with the port off (the port changes no health path); the invented-dose and blood-pressure probes stay fixed; the health predicate is byte-identical (0 new false fires); release APK 3,044,169,517 B with no INTERNET permission",
   "value": "41 fixed / 1 exposure / 23 model of 65; APK 3,044,169,517 B",
   "ci": null,
   "n": "65 probes",
   "seeds": null,
   "set": "65 F3-SAFETY probes (dev)",
   "set_status": "dev",
   "judge": "none (0 LLM)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 19:40",
   "source_public": null,
   "caveat": "Engine tests, device-test compile and debug build pass. Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-final"
   ]
  },
  {
   "id": "F-2211",
   "kind": "event",
   "claim": "BoarLite candidate contents since the first release build: health gate (B+FU detector, after retrieval, exposure rule), the fixed line 'May contain errors; check the source.' under every model answer (replacing the ported warning line), factbook v1, QuickMath, model preload, a 5-card onboarding, exposure and conversion questions answered by code with a first-aid excerpt, search port ON_NO_HEALTH; open at the build session's close: phone smoke of the final APK, APK hosting (above a 2 GB asset limit), release-key backup, About text",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 19:44",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-final-contents"
   ]
  },
  {
   "id": "F-2212",
   "kind": "fact",
   "claim": "REFS-V2, OpenAI gpt-5.6-sol half complete: 75 of 75 v0.4 references written with live web search under the frozen protocol, 0 hard failures, 97 calls (two items completed on a resumed run); references of both judge families now exist",
   "value": "75 / 75; 0 hard failures",
   "ci": null,
   "n": "75 items (88 turns)",
   "seeds": null,
   "set": "v0.4 (held-out ruler)",
   "set_status": "held-out ruler (exposed)",
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 20:05",
   "source_public": null,
   "caveat": "Protocol frozen 2026-10-21 before any call (prompt 5e41177d, generator eb2fd7e2). Reference text not read by the coordinator.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "U2-refs-codex"
   ]
  },
  {
   "id": "F-2213",
   "kind": "fact",
   "claim": "After the engine used for the final ruler run (b4ef94f3), BoarLite's main branch gained two UI-only commits (brand colours always on; an animated six-card tour over the real chat UI): 10 files, 0 engine files, so the ruler run stays valid for the final engine; the release APK is rebuilt from 6aa7508a",
   "value": "10 files changed, 0 under engine/",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 20:32",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-ui-only"
   ],
   "superseded_by": [
    "F-2223",
    "F-2225"
   ],
   "superseded_note": "2026-10-25: LabRat 7788924a changes the grounding prompt's identity line after the ruler run, so the shipped app is no longer engine-identical to b4ef94f3; the ruler number stays the number for b4ef94f3."
  },
  {
   "id": "F-2214",
   "kind": "fact",
   "claim": "Final ruler run (RULER-FINAL) on BoarLite's final engine b4ef94f3 exists as of 2026-10-24 20:34: desktop twin of the app path (QuickMath, search port ON_NO_HEALTH, health gate, history as the app sends it), 71 headline items + 4 anchors of bench v0.4, one seed, each judge family against its own references (Claude Opus vs the Claude references; OpenAI gpt-5.6-sol vs its references), 150 + 150 judge calls, addendum frozen before any generation; its numbers were held until the author's publication decision of 2026-10-25 (now public as F-2215 and F-2217; history, the headline is DEVICE-FINAL)",
   "value": null,
   "ci": null,
   "n": "71 headline items + 4 anchors",
   "seeds": "1 (42)",
   "set": "v0.4 (held-out ruler; exposed)",
   "set_status": "held-out ruler (v0.4; now exposed to three engines)",
   "judge": "Claude Opus + OpenAI gpt-5.6-sol (each vs its own references)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 20:34",
   "source_public": null,
   "caveat": "Deviations from the spec: one seed (spec: two; the author's choice), one reference family per judge family, stage 1 (DECLINE2STAGE) not run, no errata, desktop twin not the APK on a phone (the spec labels desktop-twin numbers dev/parity evidence). v0.4 is the claim set: no tuning after this run.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "U2-rf-run"
   ]
  },
  {
   "id": "F-2215",
   "kind": "fact",
   "claim": "RULER-FINAL U-total of BoarLite b4ef94f3 on v0.4: Claude family 0.164 (0.123-0.208), OpenAI gpt-5.6-sol family 0.245 (0.194-0.300); minimum over families 0.164 (0.123-0.206); win-rate 0 of 71 (Claude) and 0 of 71 with 1 tie (gpt-5.6-sol); fixed-text refusals on 3 items",
   "value": "Claude 0.164; gpt-5.6-sol 0.245; min 0.164; win-rate 0",
   "ci": "Claude 0.123-0.208; gpt-5.6-sol 0.194-0.300; min 0.123-0.206",
   "n": "71 headline items",
   "seeds": "1 (42)",
   "set": "v0.4 (held-out ruler; exposed)",
   "set_status": "held-out ruler (v0.4; exposed)",
   "judge": "Claude Opus + OpenAI gpt-5.6-sol (each vs its own references)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 20:34",
   "source_public": null,
   "caveat": "Metric: ruler v0.4 U-total (71 headline items, health weight 0.5); never on one axis with the judge.v1 ratio or outcome labels. Desktop twin, one seed. Published by the author's decision of 2026-10-25 (AD-15 (b)): BoarLite/LabRat is the author's own submission. History, not the headline: desktop twin of engine b4ef94f3, one seed, before the Rat persona line and the cleanup (F-2223, F-2225); not a measurement of LabRat 7788924a. The headline is the on-phone run DEVICE-FINAL (OF-8).",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "U2-rf-utotal"
   ]
  },
  {
   "id": "F-2217",
   "kind": "fact",
   "claim": "RULER-FINAL information-available split: Claude family U-total 0.164 where the installed packs hold the answer (n 19), 0.174 partly (n 25), 0.155 not at all (n 27); gpt-5.6-sol family 0.272 / 0.238 / 0.233: with the information present, the 1.7B does not deliver it",
   "value": "AVAILABLE 0.164 vs NONE 0.155 (Claude)",
   "ci": null,
   "n": "19 / 25 / 27 headline items",
   "seeds": "1 (42)",
   "set": "v0.4 (held-out ruler; exposed)",
   "set_status": "held-out ruler (v0.4; exposed)",
   "judge": "Claude Opus + OpenAI gpt-5.6-sol",
   "date": "2026-10-24",
   "date_text": "2026-10-24 20:34",
   "source_public": null,
   "caveat": "Metric: ruler v0.4 U-total (71 headline items, health weight 0.5); never on one axis with the judge.v1 ratio or outcome labels. Pre-registered stratifier (spec r4). Published by the author's decision of 2026-10-25 (AD-15 (b)): BoarLite/LabRat is the author's own submission. History, not the headline: desktop twin of engine b4ef94f3, one seed, before the Rat persona line and the cleanup (F-2223, F-2225); not a measurement of LabRat 7788924a. The headline is the on-phone run DEVICE-FINAL (OF-8).",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "new-ruler",
   "aliases": [
    "U2-rf-factavail"
   ]
  },
  {
   "id": "F-2218",
   "kind": "fact",
   "claim": "Finding from the final ruler run: BoarLite's chat shows the streamed model text, so the pipeline's cleaned answer (the citation support check CT-1 and attribution) is not used on the plain path; the run scored what the chat shows. Logged as a possible BoarLite bug",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 20:34",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U2-rf-ct1"
   ]
  },
  {
   "id": "F-2219",
   "kind": "fact",
   "claim": "Final BoarLite release APK (build 6aa7508a = engine b4ef94f3 plus UI-only commits: brand colours, animated six-card tour) smoke-tested on a Pixel 8a in airplane mode: fresh install, tour renders, 0 bytes of network traffic, 0 crashes, peak PSS 3.23 GB; a cited answer in 14.9 s at 16.3 tok/s and another in 9.4 s; the child-dose question gets the fixed health text; temperature conversion computed by code",
   "value": "PASS; 0 B network; 0 crashes; PSS 3.23 GB; 14.9 s at 16.3 tok/s",
   "ci": null,
   "n": "6 smoke questions",
   "seeds": null,
   "set": "device smoke (dev)",
   "set_status": "dev",
   "judge": "none (smoke checks)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 20:47",
   "source_public": null,
   "caveat": "A smoke test, not a quality measurement. One answer (minor burn) came from a first-aid source without a citation marker plus 2 unrelated sources. Health content: AI-reviewed, not clinically reviewed.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U3-apk-final-smoke"
   ],
   "superseded_by": [
    "F-2235"
   ],
   "superseded_note": "2026-10-25: later APKs (cleanup ff50c5ff, LabRat 7788924a, logo 0ad476ca) replaced the 6aa7508a release APK on the phone."
  },
  {
   "id": "F-2220",
   "kind": "event",
   "claim": "BoarLite screens review by the author (2026-10-24 evening, UI only, engine untouched so the ruler run on b4ef94f3 still applies): a calmer first tour card, Knowledge pack icons, time and tok/s shown above the sources, no Free memory switch (always on), no PIN lock (code removed), haptics on by default, a new About page, reset texts without PIN, the chat top bar showing only the name, a drawer brand card counting the 13 knowledge packs, and the mark in the accent colour. Phone smoke of the resulting APK (b45f3aa8) on the Pixel 8a: a cited answer in 24 s, the child-ibuprofen question gets the fixed health text, 40 °C = 104 °F computed, Large text checked on all main screens",
   "value": "UI only; smoke on the Pixel 8a passed",
   "ci": null,
   "n": "hand smoke",
   "seeds": null,
   "set": "device smoke (dev)",
   "set_status": "dev",
   "judge": "none (hand checks)",
   "date": "2026-10-24",
   "date_text": "2026-10-24 23:48",
   "source_public": null,
   "caveat": "A smoke test, not a quality measurement. Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech"
   ],
   "chapter": "submission",
   "aliases": [
    "U4-screens-review"
   ]
  },
  {
   "id": "F-2221",
   "kind": "decision",
   "claim": "Long-chat summary removed from BoarLite: on the Pixel the 1.7B model used only the last 3 exchanges of a long chat, so the summary was switched off (22:10) and its code removed (owner decision, commit d14804bf)",
   "value": "summary off; code removed",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 22:10",
   "source_public": null,
   "caveat": "Observation on one phone, not a measured arm.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U4-summary-off"
   ]
  },
  {
   "id": "F-2222",
   "kind": "event",
   "claim": "Product cleanup round on BoarLite (2026-10-24 22:55 → 2026-10-25 01:11, owner decisions): the Expo/React Native tree removed (the app ships the native Android code only); code unreachable in the closed app removed (Deep Research beyond the recap, the Detailed style, persona and length options, adaptive routing, the long-chat summary, PIN and sealed-chat legacy); Portuguese support removed (English only, except words that are also English); Settings without the Knowledge section or the title switch (titles always on), Clear chat history moved under Start over, 50 chats kept instead of 10; MIT licence texts shipped in About; docs describe the closed app. Engine unit tests: 274, all passing; each step device-checked on the Pixel 8a",
   "value": "main ff50c5ff; 274 engine tests",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "date_text": "2026-10-25 01:11",
   "source_public": null,
   "caveat": "Product changes; no ruler rerun after them (see the persona-line row).",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U4-cleanup"
   ]
  },
  {
   "id": "F-2223",
   "kind": "fact",
   "claim": "Between the ruler engine b4ef94f3 and LabRat 7788924a the engine module changed in 124 files (643 lines added, 97,537 removed), almost all removals of code and test resources described as unreachable in the closed app; the frozen DEVICE-FINAL criterion describes 7788924a's engine as b4ef94f3 plus the Rat persona line. Behaviour checks after the cleanup: 274 engine unit tests and a 5-question device smoke whose retrieved titles, health gate and conversion match the baseline; no ruler rerun",
   "value": "124 files; +643 / −97,537 lines",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "date_text": "2026-10-25 02:20",
   "source_public": null,
   "caveat": "Line counts from a read-only git diff of the engine module; the smoke compares titles, gate and conversion, not answer quality.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U4-engine-diff"
   ]
  },
  {
   "id": "F-2224",
   "kind": "decision",
   "claim": "BoarLite renamed LabRat at the app maintainer's request (commit 7788924a): the name changes in the UI, docs and assets; the assistant is called Rat (answer byline, notification, and the grounding prompt's identity line); a new rat mark in the app, launcher and splash. Unchanged: the package id, the code packages, the bundled knowledge catalogue and the About credit \"Based on BOAR (MIT)\". Phone smoke 6/6: titles, health gate and conversion match the baseline; \"Who are you?\" answers \"I am Rat…\"",
   "value": "BoarLite → LabRat; assistant Rat; smoke 6/6",
   "ci": null,
   "n": "6 smoke questions",
   "seeds": null,
   "set": "device smoke (dev)",
   "set_status": "dev",
   "judge": "none (smoke checks)",
   "date": "2026-10-25",
   "date_text": "2026-10-25 02:20",
   "source_public": null,
   "caveat": "A rename plus one prompt line; a smoke test, not a quality measurement. The final ruler number (RULER-FINAL) measures engine b4ef94f3 on its desktop twin; LabRat 7788924a adds the Rat persona line to the grounding prompt after that run, so the ruler number is not a measurement of LabRat 7788924a.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U4-labrat"
   ]
  },
  {
   "id": "F-2225",
   "kind": "fact",
   "claim": "Engine change after the ruler run: in LabRat 7788924a the grounding prompt's identity line changed from \"You are BOAR, an offline AI research app running on this phone…\" to \"You are Rat, the assistant in LabRat, an offline AI research app running on this phone…\". The final ruler run (RULER-FINAL, 2026-10-24 20:34) was on b4ef94f3, before this change: its number is for b4ef94f3 (desktop twin), and DEVICE-FINAL measures LabRat 7788924a on the phone",
   "value": "persona line changed after RULER-FINAL",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "date_text": "2026-10-25 02:20",
   "source_public": null,
   "caveat": "The final ruler number (RULER-FINAL) measures engine b4ef94f3 on its desktop twin; LabRat 7788924a adds the Rat persona line to the grounding prompt after that run, so the ruler number is not a measurement of LabRat 7788924a.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U4-persona-line"
   ]
  },
  {
   "id": "F-2226",
   "kind": "decision",
   "claim": "LabRat logo n8 chosen by the author after 4 logo rounds: a magnifier whose ring is a rat's head and whose handle is its tail (commit 0ad476ca); its APK installed on the Pixel 8a, smoke 6/6 matching the baseline, app screen and launcher icon checked on the phone",
   "value": "logo n8; smoke 6/6",
   "ci": null,
   "n": "6 smoke questions",
   "seeds": null,
   "set": "device smoke (dev)",
   "set_status": "dev",
   "judge": "none (smoke checks)",
   "date": "2026-10-25",
   "date_text": "2026-10-25 11:24",
   "source_public": null,
   "caveat": "A smoke test, not a quality measurement.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech"
   ],
   "chapter": "submission",
   "aliases": [
    "U4-logo-n8"
   ]
  },
  {
   "id": "F-2227",
   "kind": "fact",
   "claim": "DEVICE-FINAL criterion frozen (sha256 recorded) on 2026-10-25 03:3x, before any matrix generation: LabRat 7788924a answers on a Pixel 8a 8 GB in airplane mode on AC power. Matrix: M1 ruler v0.4 (75 items, 88 turns) on seeds 42, 43 and 44, judged as in RULER-FINAL (Claude Opus vs Claude references on all three seeds; OpenAI gpt-5.6-sol vs its references on seed 42, further seeds optional), reporting per family the U-total, CI, win-rate, refusals, the minimum over families, the three-seed mean and a phone-vs-desktop parity read against the b4ef94f3 desktop twin (not a verdict); M2 the 65 safety probes (0 LLM); M3 the 140 on-K13 pack self-tests and M4 the 60 UXMATRIX items (Claude Haiku labels); M5 latency, time to first token, decode speed, memory, battery temperature, three cold starts and the app's network counters over the night. Claim numbers = M1 (both families named, seeds stated, win-rate first) and M5 (speed, memory, offline); M2-M4 are product numbers. No tuning, no engine change",
   "value": "matrix M1-M5; claim numbers = M1 + M5",
   "ci": null,
   "n": "75 items × 3 seeds + 265 product items",
   "seeds": "42, 43, 44 (M1); 42 (M2-M4)",
   "set": "bench v0.4 (exposed); 65 safety probes; 140 on-K13 self-tests; UXMATRIX 60",
   "set_status": "dev",
   "judge": "Claude Opus + OpenAI gpt-5.6-sol (M1); Claude Haiku 5.5-low labels (M3, M4); none (M2, M5)",
   "date": "2026-10-25",
   "date_text": "2026-10-25 03:3x",
   "source_public": null,
   "caveat": "No DEVICE-FINAL result is in this record; the matrix was running as of this update.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U4-df-criterion"
   ]
  },
  {
   "id": "F-2228",
   "kind": "event",
   "claim": "DEVICE-FINAL started overnight on 2026-10-25 at the author's request (final benchmarks with the app running the engine on the phone, for publication): the device eval runner was patched on an eval branch to answer follow-up turns with the chat's history and to record the streamed text the chat shows; engine logic untouched; the unbundled release was installed over the bundled data with the same signing identity; a pre-run smoke and a follow-up-turn check passed",
   "value": "running",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "date_text": "2026-10-25 03:3x",
   "source_public": null,
   "caveat": "No DEVICE-FINAL result is in this record; the matrix was running as of this update.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U4-df-start"
   ]
  },
  {
   "id": "F-2229",
   "kind": "event",
   "claim": "An APK install on the Pixel 8a interrupted the DEVICE-FINAL run on 2026-10-25 11:28; the run was restarted, and the phone is kept free of installs until the matrix log ends",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "date_text": "2026-10-25 11:28",
   "source_public": null,
   "caveat": "No DEVICE-FINAL result is in this record; the matrix was running as of this update. Process fact only.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U4-df-interrupt"
   ],
   "superseded_by": [
    "F-2244"
   ],
   "superseded_note": "2026-10-25 final: the run was resumed (69 complete items kept, 6 re-run), not restarted from scratch."
  },
  {
   "id": "F-2230",
   "kind": "decision",
   "claim": "APK hosting decided: the 3.04 GB LabRat APK goes to Hugging Face (author's OK on 2026-10-25) because GitHub release assets must be under 2 GiB; a public install page is planned with the download link, sha256, about 6 GB free space, Android 8+, \"install unknown apps\", the Play Protect prompt and a first launch of about 12 s. The code repository stays private unless the author decides otherwise",
   "value": "Hugging Face; upload pending",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "source_public": null,
   "caveat": "As of this update the upload is not recorded as done.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U4-hosting"
   ],
   "superseded_by": [
    "F-2246"
   ],
   "superseded_note": "2026-10-25 15:3x: the APK is hosted on Hugging Face (huggingface.co/r4topunk/labrat); the code repository is public (labrat-boar-fork)."
  },
  {
   "id": "F-2231",
   "kind": "fact",
   "claim": "Install-path limits found before release: Android developer verification is live since 2026-09-30 in Brazil, Indonesia, Singapore and Thailand; adb installs are exempt, but browser installs of apps from unregistered developers may hit an \"advanced flow\" (a warning and a one-day wait), scope unconfirmed. The test phone runs GrapheneOS, which is not Google-certified, so it cannot prove the judges' install path; a browser-install test on a stock certified phone is planned, and registering the app with the developer verification is the mitigation (needs the author's Google account)",
   "value": "risk open",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "source_public": null,
   "caveat": "Policy scope as checked by the build session on 2026-10-25, not confirmed on a certified phone.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U4-install-risk"
   ]
  },
  {
   "id": "F-2232",
   "kind": "fact",
   "claim": "Open engine bug, as of 2026-10-25: the citation-support step (Citations.applyShown) is not applied to the streamed answer the chat shows (found by the final ruler run). Fixing it changes the engine and would need a ruler rerun; the decision is the author's. RULER-FINAL scored, and DEVICE-FINAL records, what the chat shows",
   "value": "open; fix needs a ruler rerun",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "source_public": null,
   "caveat": "No fix is in the record as of this update.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U4-applyshown"
   ]
  },
  {
   "id": "F-2233",
   "kind": "fact",
   "claim": "Claim-text plan (2026-10-24 PM hand-off): lead with what the app does and the final ruler numbers labelled with both judge families, the seed count, \"desktop twin\" and the win-rate, plus measured product numbers each labelled with the engine they come from; wording goes to an independent reviewer, then the author. Also pending: a decisions-and-evidence page for the judges, a short demo post and the app's About text",
   "value": "plan; not posted",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 20:5x",
   "source_public": null,
   "caveat": "The ruler numbers stay internal until the author rules on AD-15; this row states the plan, not the numbers.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "U4-claim-plan"
   ]
  },
  {
   "id": "F-2234",
   "kind": "fact",
   "claim": "Post-submission backlog (not before the submission): medical Wikipedia titles as health sources; a dose guard (cut ungrounded or toxic-threshold doses) so the search port can run on health turns; X2B-BUDGET 900 and RETBENCH k4+comp40 (both passed the grounded-dose re-gate) into the native search port; factbook v2 with the new search",
   "value": null,
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-24",
   "date_text": "2026-10-24 20:5x",
   "source_public": null,
   "caveat": "Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "handoff",
   "aliases": [
    "U4-backlog"
   ]
  },
  {
   "id": "F-2235",
   "kind": "fact",
   "claim": "Latest LabRat APK (0ad476ca, bundled, same signing identity, no INTERNET permission) installed on the Pixel 8a after the rename and logo: smoke 6/6 matching the baseline (retrieved titles, health gate and conversion); the earlier cleanup build ff50c5ff is 3.04 GB. The APK built from b4ef94f3, the ruler engine, is kept",
   "value": "smoke 6/6",
   "ci": null,
   "n": "6 smoke questions",
   "seeds": null,
   "set": "device smoke (dev)",
   "set_status": "dev",
   "judge": "none (smoke checks)",
   "date": "2026-10-25",
   "date_text": "2026-10-25 11:24",
   "source_public": null,
   "caveat": "A smoke test, not a quality measurement. The final ruler number (RULER-FINAL) measures engine b4ef94f3 on its desktop twin; LabRat 7788924a adds the Rat persona line to the grounding prompt after that run, so the ruler number is not a measurement of LabRat 7788924a.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech"
   ],
   "chapter": "submission",
   "aliases": [
    "U4-labrat-apk"
   ]
  },
  {
   "id": "F-2236",
   "kind": "decision",
   "claim": "The author restated the thesis on 2026-10-25: small models need external knowledge; the project learned that and built a harness around a small model's real capabilities, tuned for the app user's experience (health gate, QuickMath, pack search and delivery, the Succinct persona, a fixed check-the-source line, refusal behaviour). It replaces the 2026-10-23 sentence \"right whenever the content is in a pack\" as the thesis statement; the conditional closing-test numbers are its evidence and its limits",
   "value": "thesis statement (author, 2026-10-25)",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "date_text": "2026-10-25",
   "source_public": null,
   "caveat": "who: author; a framing statement, not a measurement. Evidence: with the matching pack the 1.7B is right on 48 % vs 26 % of pack-written self-tests (F-1704); 65 % when the item's own article reaches the model vs 28 % (F-1705). Limits: the article is delivered on 54 % (F-1706); no pack reached the 0.75 bar (F-1707); no gain on real development items (F-2080). Never \"the harness makes it half as good\".",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "closure",
   "aliases": [
    "U5-thesis-2026-10-25"
   ]
  },
  {
   "id": "F-2237",
   "kind": "decision",
   "claim": "Author publication decisions of 2026-10-25: numbers of the author's own submission (BoarLite, renamed LabRat) are publishable, numbers of the app maintainer's app stay internal; the headline is the final v0.4 ruler measured on the phone (DEVICE-FINAL); the desktop ruler run of b4ef94f3 is history with its code-difference note; the citation-support gap (Citations.applyShown) is documented as a known limitation and not fixed before the measurement",
   "value": "LabRat numbers public; headline = DEVICE-FINAL; applyShown not fixed",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "date_text": "2026-10-25",
   "source_public": null,
   "caveat": "who: author. Not fixing applyShown keeps the measured engine identical to the app judges install; the app maintainer's app numbers need his consent.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U5-pubdecisions-2026-10-25"
   ]
  },
  {
   "id": "F-2238",
   "kind": "fact",
   "claim": "DEVICE-FINAL M1, the final ruler measured on the phone: LabRat on a Pixel 8a (8 GB, airplane mode) answered bench v0.4 (71 headline items + 4 anchors) on seeds 42 and 43. It won 0 of 71 head-to-heads under both judge families (one tie under OpenAI gpt-5.6-sol on seed 42). U-total, both seeds pooled: Claude Opus 0.165 (0.127–0.205), OpenAI gpt-5.6-sol 0.242 (0.194–0.296). Per seed: Claude 0.167 (0.129–0.209) and 0.162 (0.125–0.201); gpt-5.6-sol 0.244 (0.193–0.298) and 0.241 (0.190–0.297). The lower family, Claude at 0.165 (0.127–0.205), is below the 0.50 bar. 5 headline items were refusals; answer coverage 0.90",
   "value": "0 of 71 wins under both families; U-total 0.165 (Claude) / 0.242 (gpt-5.6-sol); below the 0.50 bar",
   "ci": "Claude 0.127-0.205; gpt-5.6-sol 0.194-0.296 (pooled seeds 42+43)",
   "n": "71 headline items + 4 anchors",
   "seeds": "42, 43",
   "set": "bench v0.4 (exposed)",
   "set_status": "held-out ruler (v0.4; exposed)",
   "judge": "Claude Opus + OpenAI gpt-5.6-sol via the Codex CLI (each vs its own references)",
   "date": "2026-10-25",
   "date_text": "2026-10-25 13:2x",
   "source_public": null,
   "caveat": "Metric: ruler v0.4 U-total (71 headline items, health weight 0.5); never on one axis with the judge.v1 ratio or outcome labels. v0.4 has been answered by our engines since 2026-10-24, so it is no longer unseen (F-2214). One reference family per judge family; stage 1 not run; no errata. Two seeds (42, 43): the criterion listed 42, 43 and 44; seed 44 was dropped for time (two seeds = the ruler spec). Measured build: LabRat 0ad476ca (engine = 7788924a, b4ef94f3 plus the Rat persona line; 0ad476ca changes brand files only). The released LabRat 1.0.0 (dcc515fc) has the same engine module (empty git diff); its later changes are UI, credits, docs and build scripts.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "U6-df-m1"
   ]
  },
  {
   "id": "F-2239",
   "kind": "fact",
   "claim": "DEVICE-FINAL phone vs desktop twin (seed 42, paired over the 71 headline items): Claude 0.167 on the phone vs 0.164 for the desktop twin of b4ef94f3, +0.003 (−0.022…+0.026), 21 items better and 16 worse; gpt-5.6-sol 0.244 vs 0.245, −0.007 (−0.039…+0.023), 22 better and 22 worse. The phone repeats the desktop number; a parity read, not a verdict",
   "value": "paired +0.003 (Claude) / −0.007 (gpt-5.6-sol)",
   "ci": "Claude −0.022…+0.026; gpt-5.6-sol −0.039…+0.023",
   "n": "71 headline items",
   "seeds": "42",
   "set": "bench v0.4 (exposed)",
   "set_status": "held-out ruler (v0.4; exposed)",
   "judge": "Claude Opus + OpenAI gpt-5.6-sol via the Codex CLI (each vs its own references)",
   "date": "2026-10-25",
   "date_text": "2026-10-25 13:2x",
   "source_public": null,
   "caveat": "Metric: ruler v0.4 U-total (71 headline items, health weight 0.5); never on one axis with the judge.v1 ratio or outcome labels. v0.4 has been answered by our engines since 2026-10-24, so it is no longer unseen (F-2214). One reference family per judge family; stage 1 not run; no errata. The two engines differ by the Rat persona line and platform sampling. Measured build: LabRat 0ad476ca (engine = 7788924a, b4ef94f3 plus the Rat persona line; 0ad476ca changes brand files only). The released LabRat 1.0.0 (dcc515fc) has the same engine module (empty git diff); its later changes are UI, credits, docs and build scripts.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "honest-number",
   "aliases": [
    "U6-df-parity"
   ]
  },
  {
   "id": "F-2240",
   "kind": "fact",
   "claim": "DEVICE-FINAL M5, speed, memory and offline on the Pixel 8a (model-written turns of v0.4 seed 42, n 82): total answer time p50 13.9 s (p90 19.6 s), first word p50 9.8 s (p90 14.5 s), decode 11.8 tokens/s, search 1.6 s, peak memory (RSS) 3.87 GB, battery at most 36.4 °C. After 45 minutes of continuous use (seed 43): 17.0 s p50 and 8.9 tokens/s. Cold starts: model load 1.6–1.7 s, first word 10.4–19.4 s. A fresh install on a cool phone decoded 16.7 tokens/s. Network: the app has no traffic record in the phone's network statistics before or after the matrix, and no internet permission",
   "value": "p50 13.9 s; first word 9.8 s; 11.8 tok/s; 3.87 GB; 0 network traffic",
   "ci": null,
   "n": "82 model-written turns per seed",
   "seeds": "42 (43 for the warm read)",
   "set": "bench v0.4 turns on the phone",
   "set_status": "dev",
   "judge": "none (device metrics)",
   "date": "2026-10-25",
   "date_text": "2026-10-25 13:2x",
   "source_public": null,
   "caveat": "One phone (Pixel 8a, GrapheneOS, AC power, screen on). Heat roughly halves decode speed under sustained use: quote the range. Measured build: LabRat 0ad476ca (engine = 7788924a, b4ef94f3 plus the Rat persona line; 0ad476ca changes brand files only). The released LabRat 1.0.0 (dcc515fc) has the same engine module (empty git diff); its later changes are UI, credits, docs and build scripts.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U6-df-m5"
   ]
  },
  {
   "id": "F-2241",
   "kind": "fact",
   "claim": "DEVICE-FINAL M2, the 65 written health and safety probes on the phone (0 LLM scoring): 53 got the fixed safe health text and 12 went to the model (the desktop twin of b4ef94f3: 41 fixed, 1 exposure answer, 23 model); 2 model answers added a dose figure the question did not ask for. One of them (a carbon-monoxide reading of 200 ppm at home) calls the level unsafe but understates an emergency; the author kept it as a known limitation",
   "value": "53 of 65 fixed text; 2 new dose figures",
   "ci": null,
   "n": "65 probes",
   "seeds": "42",
   "set": "65 written safety probes (dev; the gate's rules were tuned on them)",
   "set_status": "dev",
   "judge": "none (0 LLM scorer)",
   "date": "2026-10-25",
   "date_text": "2026-10-25 13:2x",
   "source_public": null,
   "caveat": "Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice. Measured build: LabRat 0ad476ca (engine = 7788924a, b4ef94f3 plus the Rat persona line; 0ad476ca changes brand files only). The released LabRat 1.0.0 (dcc515fc) has the same engine module (empty git diff); its later changes are UI, credits, docs and build scripts.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U6-df-m2"
   ]
  },
  {
   "id": "F-2242",
   "kind": "fact",
   "claim": "DEVICE-FINAL M3, 140 pack-written self-tests of the 7 tested packs that LabRat ships (20 each) on the phone: correct-any 0.400 (0.314–0.479); 0.476 when the question's own article was kept (n 82) vs 0.293 when it was not (n 58); wrong-any 0.436",
   "value": "correct 0.40 of 140",
   "ci": "0.314-0.479",
   "n": "140",
   "seeds": "42",
   "set": "pack-written self-tests (dev)",
   "set_status": "dev (pack-written self-tests)",
   "judge": "Claude Haiku 5.5-low labels (UXMATRIX protocol)",
   "date": "2026-10-25",
   "date_text": "2026-10-25 13:2x",
   "source_public": null,
   "caveat": "Outcome-label shares, not the ruler scale. Self-tests favour the packs. Correct-with-source is a lower bound (phone rows hold source titles only). Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U6-df-m3"
   ]
  },
  {
   "id": "F-2243",
   "kind": "fact",
   "claim": "DEVICE-FINAL M4, the 60 UXMATRIX items on the phone: correct-any 0.350 (0.233–0.467), wrong-any 0.617",
   "value": "correct 0.35 of 60",
   "ci": "0.233-0.467",
   "n": "60",
   "seeds": "42",
   "set": "UXMATRIX 60 (dev)",
   "set_status": "dev",
   "judge": "Claude Haiku 5.5-low labels (UXMATRIX protocol)",
   "date": "2026-10-25",
   "date_text": "2026-10-25 13:2x",
   "source_public": null,
   "caveat": "Outcome-label shares, not the ruler scale. Correct-with-source is a lower bound (phone rows hold source titles only).",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U6-df-m4"
   ]
  },
  {
   "id": "F-2244",
   "kind": "event",
   "claim": "DEVICE-FINAL ran as one interrupted and resumed run: seed 42 stopped at 77 of 88 turns at 11:31 when an APK install of LabRat 0ad476ca (brand files only) reached the phone; the 69 complete items were kept and the 6 unfinished ones were re-run on 0ad476ca; seeds 43 and the M2-M5 matrix ran entirely on 0ad476ca; seed 44 was dropped for time",
   "value": "interrupted once, resumed; seed 44 dropped",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "date_text": "2026-10-25 11:31",
   "source_public": null,
   "caveat": "Process fact.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U6-df-process"
   ]
  },
  {
   "id": "F-2245",
   "kind": "decision",
   "claim": "Author decision (2026-10-25 14:0x): the carbon-monoxide probe answer stays a known limitation; no health-gate change before the submission",
   "value": "known limitation; no gate change",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "date_text": "2026-10-25 14:0x",
   "source_public": null,
   "caveat": "who: author. Health content: AI-reviewed, not clinically reviewed. Nothing here is medical advice.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U6-d-co"
   ]
  },
  {
   "id": "F-2246",
   "kind": "event",
   "claim": "LabRat 1.0.0 released: APK dcc515fc, 3,041,570,902 bytes, sha256 d2e00ba4a9adff6f903eab7d34759fd4f4db9cd721303dc7c38ae280b465dae8, hosted on Hugging Face at huggingface.co/r4topunk/labrat and installed on the Pixel 8a. The bundle was rebuilt from public pinned URLs only (16 of 16 inputs fetched and checked by sha256)",
   "value": "1.0.0 hosted; sha256 d2e00ba4…",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "date_text": "2026-10-25 15:3x",
   "source_public": null,
   "caveat": "Measured build: LabRat 0ad476ca (engine = 7788924a, b4ef94f3 plus the Rat persona line; 0ad476ca changes brand files only). The released LabRat 1.0.0 (dcc515fc) has the same engine module (empty git diff); its later changes are UI, credits, docs and build scripts.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U6-labrat-final"
   ]
  },
  {
   "id": "F-2247",
   "kind": "event",
   "claim": "The LabRat source is public at github.com/r4topunk/labrat-boar-fork (MIT, a fork of BOAR), with README, results, pack provenance and licence; made public with the author's OK",
   "value": "public repository",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "date_text": "2026-10-25 15:4x",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U6-labrat-public"
   ]
  },
  {
   "id": "F-2248",
   "kind": "fact",
   "claim": "Knowledge-pack counts, one meaning each: LabRat ships 13 packs (the K13 kit), and every one of the 13 exact files is downloadable from the boar-packs dataset on Hugging Face (7 of them under labrat-k13/: 6 earlier packs with added chunk vectors and 1 new pack). The 2026-10-23 self-tests covered 9 packs (180 questions), 7 of them in the app; DEVICE-FINAL tested those 7 on the phone (140 questions)",
   "value": "13 shipped; 9 self-tested; 7 tested on the phone",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "date_text": "2026-10-25",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U6-packs-count"
   ]
  },
  {
   "id": "F-2249",
   "kind": "event",
   "claim": "Hand check of LabRat 1.0.0 with the user's own documents on the Pixel 8a: a PDF and a Markdown file imported; with 'Only my files' 3 of 3 answers correct with the file as the source; a question not in the files got the fixed not-found line; removing a document works",
   "value": "3 of 3",
   "ci": null,
   "n": "3 questions",
   "seeds": null,
   "set": "hand check (dev)",
   "set_status": "dev",
   "judge": "none (hand check)",
   "date": "2026-10-25",
   "date_text": "2026-10-25 15:5x",
   "source_public": null,
   "caveat": "A hand check, not a benchmark.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech"
   ],
   "chapter": "submission",
   "aliases": [
    "U6-labrat-docs"
   ]
  },
  {
   "id": "F-2250",
   "kind": "fact",
   "claim": "UI findings left open on the 1.0.0 candidate (83ebb269): 'Convert 30 C to F' went to the model and showed raw LaTeX, follow-up turns pulled unrelated sources, and chat titles came out in CamelCase",
   "value": "open UI findings",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "date_text": "2026-10-25 14:5x",
   "source_public": null,
   "caveat": "Hand checks on one phone.",
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U6-ui-open"
   ]
  },
  {
   "id": "F-2251",
   "kind": "event",
   "claim": "Submission post prepared: one post with a 60 s video and a short poidh text, numbers from DEVICE-FINAL; as of the last record row it was not posted",
   "value": "prepared; not posted",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "date_text": "2026-10-25 16:29",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U6-post-ready"
   ]
  },
  {
   "id": "F-2252",
   "kind": "event",
   "claim": "Article \"RatLab — A Boar Research\" published on GitHub Pages at r4topunk.github.io/boar-research (16:33), updated at 16:46 with the DEVICE-FINAL result and links to the public repositories",
   "value": "article live",
   "ci": null,
   "n": null,
   "seeds": null,
   "set": null,
   "set_status": null,
   "judge": null,
   "date": "2026-10-25",
   "date_text": "2026-10-25 16:46",
   "source_public": null,
   "caveat": null,
   "verification": "CONFIRMED",
   "visibility": "public",
   "layer_tags": [
    "curious",
    "tech",
    "ai"
   ],
   "chapter": "submission",
   "aliases": [
    "U6-article-live"
   ]
  }
 ]
}