{ "schema": "boar-publication/decisions", "schema_version": "1.0.0", "generated": "2026-10-25", "license": "CC BY 4.0 (text and data); see /CITATION.cff", "count": 208, "notes": [], "decisions": [ { "id": "D-001", "date": "2026-09-26", "time": null, "title": "Thesis: \"quality per second + blind evaluation against a frontier model\"; stay on the existing cross-platform app; plan a no-network Android submission", "context": "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", "decision": "Thesis: \"quality per second + blind evaluation against a frontier model\"; stay on the existing cross-platform app; plan a no-network Android submission", "alternatives": [ "fork a new app; rely on the existing on-phone benchmark" ], "rationale": "the existing claim had solid engineering but no measured research quality", "consequence": "led to the 09-29 research pipeline and 10-01 harness", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-050" ], "chapter": "question", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-01" ] }, { "id": "D-002", "date": "2026-09-30", "time": null, "title": "References = Claude Opus + web search, frozen once, hashed; never regenerated", "context": "Frozen frontier-plus-web reference answers for the public v0.1 set start being generated (09-30 → 10-02)", "decision": "References = Claude Opus + web search, frozen once, hashed; never regenerated", "alternatives": [ "live references per run" ], "rationale": "a fixed denominator makes runs comparable", "consequence": "same refs used for every v0.1 number; regenerating = not comparable", "reversible": "no (by design)", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-059" ], "chapter": "question", "visibility": "public", "status_label": "kept", "aliases": [ "D-method-04" ] }, { "id": "D-003", "date": "2026-10-01", "time": null, "title": "Build one evaluation harness (ADR-001): own thin protocol + existing scorers; 5 layers; 7 axes never averaged", "context": "ADR-001 proposes one evaluation harness: 5 layers, one trace schema, 7 axes never averaged, gate vs full modes", "decision": "Build one evaluation harness (ADR-001): own thin protocol + existing scorers; 5 layers; 7 axes never averaged", "alternatives": [ "Inspect AI (UK AISI); promptfoo", "Inspect AI (Python, subprocess bridge); promptfoo (no paired bootstrap)" ], "rationale": "each experiment wrote its own scripts; desktop and device rows had different shapes; no regression gate; run noise ≈ effect size", "consequence": "harness became boar-eval, later the public boar-bench", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1305", "F-678" ], "chapter": "phone", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-02", "D-method-01" ] }, { "id": "D-004", "date": "2026-10-01", "time": null, "title": "Replace the 27 items borrowed from a competitor's question list with own items (bench v0.1)", "context": "27 items from a competing team's fork removed from the bench; replaced with original items of the same slot (bench v0.1)", "decision": "Replace the 27 items borrowed from a competitor's question list with own items (bench v0.1)", "alternatives": [ "keep them", "keep them with attribution" ], "rationale": "do not copy competitors' test cases; clean benchmark", "consequence": "v0.1 = 60 public items (57 EN judged)", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-099", "F-101" ], "chapter": "ruler", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-03", "D-method-03" ] }, { "id": "D-005", "date": "2026-10-01", "time": "23:00", "title": "Report a quality effect only if it exceeds the MDE printed by the diff tool", "context": "Rule: an effect is reported only if it exceeds the minimum detectable effect (MDE)", "decision": "Report a quality effect only if it exceeds the MDE printed by the diff tool", "alternatives": [ "report any sign" ], "rationale": "run-to-run noise (Δ 0.031) as large as effects", "consequence": "became house rule; later refined (two-seed, seed-noise rule)", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1137" ], "chapter": "measurer", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-04" ] }, { "id": "D-006", "date": "2026-10-01", "time": null, "title": "Freeze every deciding criterion (sha256) before data; blind sets written by separate writers, never read by researchers", "context": "Embedding danger detector criterion frozen before its blind set exists (start of learned-detector line)", "decision": "Freeze every deciding criterion (sha256) before data; blind sets written by separate writers, never read by researchers", "alternatives": [ "iterate without freezes" ], "rationale": "avoid post-hoc tuning", "consequence": "65/65 criterion shas verify by 10-10", "reversible": "yes (method)", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-100" ], "chapter": "ruler", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-05" ] }, { "id": "D-007", "date": "2026-10-01", "time": null, "title": "Seven axes (quality, truthfulness, safety, retrieval, latency, resources, stability), never averaged into one score", "context": "Research mandate: three researcher lanes (retrieval, models, evaluation/safety) write \"technique cards\"; a runner executes arms", "decision": "Seven axes (quality, truthfulness, safety, retrieval, latency, resources, stability), never averaged into one score", "alternatives": [ "single composite score" ], "rationale": "one average hides safety and latency regressions", "consequence": "quality ratio became the published number; other axes reported separately", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-290" ], "chapter": "builds", "visibility": "public", "status_label": "kept", "aliases": [ "D-method-02" ] }, { "id": "D-008", "date": "2026-10-01", "time": null, "title": "Research agents work in fixed lanes, write one \"technique card\" per idea, never run experiments or touch the product; only the coordinator queues runs", "context": "Research mandate: three researcher lanes (retrieval, models, evaluation/safety) write \"technique cards\"; a runner executes arms", "decision": "Research agents work in fixed lanes, write one \"technique card\" per idea, never run experiments or touch the product; only the coordinator queues runs", "alternatives": [ "Agents run their own tests" ], "rationale": "Avoid duplicate work; separate proposing from measuring", "consequence": "the retrieval research agent 19 cards, the models/latency research agent 22, the evaluation/safety research agent 23; runner agent executed", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-823" ], "chapter": "agent-lab", "visibility": "public", "status_label": "kept", "aliases": [ "D-process-01" ] }, { "id": "D-009", "date": "2026-10-01", "time": null, "title": "Every pass/fail criterion written and sha256-frozen before the run; no edits after verdicts, only dated addenda", "context": "First frozen criteria on fresh sets (hm-fresh2/3) before any reading", "decision": "Every pass/fail criterion written and sha256-frozen before the run; no edits after verdicts, only dated addenda", "alternatives": [ "Judge then decide" ], "rationale": "Prevent tuning the rule to the result", "consequence": "65/65 criterion hashes verified 10-10; 16 of 17 re-verified by an outside critic 10-16", "reversible": "no (core method)", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": "Verification note: \"18/19 re-verified\" → 16/17 (CRITIQUE:6) or 19/20 in the C6 spot-check (CRITIQUE:72-75).", "fact_ids": [ "F-1136" ], "chapter": "measurer", "visibility": "public", "status_label": "kept", "aliases": [ "D-process-02" ] }, { "id": "D-010", "date": "2026-10-02", "time": "06:10", "title": "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, perso", "context": "Strategy memo (high-tier model, one call): Android-only, three fronts; stop iPhone/1.5B and prompt work", "decision": "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", "alternatives": [ "continue broad fronts" ], "rationale": "measured gap; only the 4B near the bar", "consequence": "iPhone work stopped; knowledge front closed later that day", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-697" ], "chapter": "phone", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-06" ] }, { "id": "D-011", "date": "2026-10-02", "time": "11:15", "title": "Bar on the RAW ratio; every publication shows raw + floor-removed + above-always-decline side by side", "context": "Author's morning decisions: bar on the raw ratio, but always publish raw + floor-removed + above-decline side by side", "decision": "Bar on the RAW ratio; every publication shows raw + floor-removed + above-always-decline side by side", "alternatives": [ "lead with floor-adjusted", "Floor-adjusted only" ], "rationale": "rubric floor gives a blank system ~0.20–0.31", "consequence": "headline later switched to floor-adjusted (D-chron-early-21) while raw stays the formal rule", "reversible": "yes", "status": "superseded", "status_note": "Floor-adjusted ratio + win/tie/loss became the headline on 2026-10-06 (applied 2026-10-07); the raw rule is still printed as the formal rule.", "superseded_by": "D-036", "verification_note": null, "fact_ids": [ "F-264", "F-271" ], "chapter": "floor-trap", "visibility": "public", "status_label": "superseded by D-036", "aliases": [ "D-chron-early-07", "D-process-05" ] }, { "id": "D-012", "date": "2026-10-02", "time": "12:40", "title": "Close the knowledge-pack front for the bounty after the big-pack gate failed replay (\"no fourth pack iteration\")", "context": "Big-pack relevance gate fails replay → knowledge front closed for the bounty (kill rule K1)", "decision": "Close the knowledge-pack front for the bounty after the big-pack gate failed replay (\"no fourth pack iteration\")", "alternatives": [ "a fourth pack iteration (G1′)" ], "rationale": "kill rule K1 pre-registered in the strategy memo", "consequence": "packs reopened later only for the deep tier and as products", "reversible": "yes", "status": "superseded", "status_note": "packs reopened later only for the deep tier and as products", "superseded_by": null, "verification_note": null, "fact_ids": [ "F-072" ], "chapter": "question", "visibility": "public", "status_label": "superseded", "aliases": [ "D-chron-early-08" ] }, { "id": "D-013", "date": "2026-10-02", "time": "12:55", "title": "Run the deep tier (35B MoE); its ratio counts only reported separately from the default, with its latency", "context": "Deep tier (35B MoE) approved; counts only if reported separately from the default", "decision": "Run the deep tier (35B MoE); its ratio counts only reported separately from the default, with its latency", "alternatives": [ "ignore deep tier" ], "rationale": "first candidate lever with a prior above noise; bounty hint about large MoE", "consequence": "deep35 +0.060 (first lever above MDE)", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-074" ], "chapter": "question", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-09" ] }, { "id": "D-014", "date": "2026-10-02", "time": null, "title": "Treat the bounty as qualitative: judged by a public confirmation, not a numeric benchmark", "context": "Bounty re-read: winning is a public confirmation, not a numeric benchmark", "decision": "Treat the bounty as qualitative: judged by a public confirmation, not a numeric benchmark", "alternatives": [ "treat our ratio as the decision" ], "rationale": "bounty text", "consequence": "private ratio kept as instrument; claim framing open", "reversible": "n/a", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-111" ], "chapter": "ruler", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-10" ] }, { "id": "D-015", "date": "2026-10-02", "time": null, "title": "Hold v0.2 as held-out: develop on v0.1, confirm only on v0.2", "context": "Direction: two tracks only (harness + pipeline technique); v0.2 = held-out, confirm only", "decision": "Hold v0.2 as held-out: develop on v0.1, confirm only on v0.2", "alternatives": null, "rationale": "avoid tuning on the claim set", "consequence": "v0.2 later burnt (D-chron-early-22)", "reversible": "yes", "status": "superseded", "status_note": "v0.2 was declared burnt → dev on 2026-10-06.", "superseded_by": "D-037", "verification_note": null, "fact_ids": [ "F-112" ], "chapter": "ruler", "visibility": "public", "status_label": "superseded by D-037", "aliases": [ "D-chron-early-11" ] }, { "id": "D-016", "date": "2026-10-02", "time": null, "title": "Restate deep35 memory: \"≤ 8 GiB\" was our own estimate, not a bounty rule; macOS footprint ≠ phone RSS", "context": "Strategy memo (high-tier model, one call): Android-only, three fronts; stop iPhone/1.5B and prompt work", "decision": "Restate deep35 memory: \"≤ 8 GiB\" was our own estimate, not a bounty rule; macOS footprint ≠ phone RSS", "alternatives": [ "keep \"fails memory\"" ], "rationale": "the author asked why 8 GB if the bounty says 12 GB", "consequence": "Linux 6,750 MiB (≈ 6.6 GiB) container test passed 10/10", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": "Verification note: Same unit fix as F-chron-early-30.", "fact_ids": [ "F-069" ], "chapter": "question", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-12" ] }, { "id": "D-018", "date": "2026-10-02", "time": null, "title": "\"Iterate fast first, rigor later\": sha freeze only when an idea goes to the full bench or held-out; funnel L0 → L1 → L2 → L3", "context": "Author: \"iterate fast first, rigor later\" (sha only when an idea reaches the full/held-out bench); no fixed greeting rules", "decision": "\"Iterate fast first, rigor later\": sha freeze only when an idea goes to the full bench or held-out; funnel L0 → L1 → L2 → L3", "alternatives": [ "freeze every step" ], "rationale": "speed of iteration", "consequence": "funnel became standard (BOUNTY-BOARD rules)", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-113" ], "chapter": "ruler", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-14" ] }, { "id": "D-019", "date": "2026-10-02", "time": null, "title": "Pairwise judging in both A/B orders; disagreement = tie; scores averaged across orders", "context": "Rubric-floor risk found: a blank system scores ~0.20–0.31 of the reference", "decision": "Pairwise judging in both A/B orders; disagreement = tie; scores averaged across orders", "alternatives": [ "single order" ], "rationale": "removes position bias", "consequence": "position consistency 96.5-100 % measured", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1139" ], "chapter": "measurer", "visibility": "public", "status_label": "kept", "aliases": [ "D-method-05" ] }, { "id": "D-020", "date": "2026-10-02", "time": null, "title": "Freeze the analysis plan (metrics, CI method, seeds, wording rules) before items existed", "context": "Held-out v0.2 analysis plan frozen (combined sha fe100e91…); items frozen in a second commit", "decision": "Freeze the analysis plan (metrics, CI method, seeds, wording rules) before items existed", "alternatives": [ "decide after seeing numbers" ], "rationale": "prevents tuning metric to results", "consequence": "plan hashed (375ff004…); later changes only as dated addenda", "reversible": "no", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-114" ], "chapter": "ruler", "visibility": "public", "status_label": "kept", "aliases": [ "D-method-06" ] }, { "id": "D-021", "date": "2026-10-02", "time": null, "title": "Primary metric = raw ratio, bar > 0.50 (the author's decision on 2026-10-02, not the app maintainer's)", "context": "Direction: \"eval + pipeline technique\" (better answers from the same model)", "decision": "Primary metric = raw ratio, bar > 0.50 (the author's decision on 2026-10-02, not the app maintainer's)", "alternatives": [ "floor-aware metric" ], "rationale": null, "consequence": null, "reversible": "—", "status": "superseded", "status_note": "Floor-adjusted ratio + win/tie/loss became the headline on 2026-10-06 (applied 2026-10-07); the raw rule is still printed as the formal rule.", "superseded_by": "D-036", "verification_note": "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'", "fact_ids": [ "F-1306" ], "chapter": "reproduce", "visibility": "public", "status_label": "superseded by D-036", "aliases": [ "D-method-07" ] }, { "id": "D-022", "date": "2026-10-02", "time": null, "title": "Blind / held-out sets: researchers never open, list or grep them; coordinator runs them by script, aggregates only", "context": "Blind-set rule widened after a filename-only grep leaked set composition", "decision": "Blind / held-out sets: researchers never open, list or grep them; coordinator runs them by script, aggregates only", "alternatives": [ "Trust-based access" ], "rationale": "A filename-only grep had leaked set composition", "consequence": "Rule widened to blind/ on 10-10; pre-commit hook blocks blind plaintext", "reversible": "no", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": "Verification note: Same fix as the timeline row: incident at QUEUE:144; MANDATE scope widening dated 10-10.", "fact_ids": [ "F-828" ], "chapter": "agent-lab", "visibility": "public", "status_label": "kept", "aliases": [ "D-process-03" ] }, { "id": "D-023", "date": "2026-10-02", "time": null, "title": "Expensive strategy agent (the strategy agent, Fable tier) only for few big decisions, with a self-contained dossier; never loops", "context": "Strategy agent (the strategy agent) memo S1: three fronts, ~1 in 4 chance the CI lower bound clears 0.50; stop list", "decision": "Expensive strategy agent (the strategy agent, Fable tier) only for few big decisions, with a self-contained dossier; never loops", "alternatives": [ "Continuous strategy agent" ], "rationale": "Cost: most expensive tier", "consequence": "Used once (S1 memo) in the record read", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-829" ], "chapter": "agent-lab", "visibility": "public", "status_label": "kept", "aliases": [ "D-process-04" ] }, { "id": "D-024", "date": "2026-10-02", "time": null, "title": "Make the memory-fit estimate repack-aware and load with no_extra_bufts on Android only", "context": "iPhone 13 repack probe: llama.rn does not repack on iOS; fit-repack stays Android-only", "decision": "Make the memory-fit estimate repack-aware and load with no_extra_bufts on Android only", "alternatives": [ "apply to iOS too; leave as is" ], "rationale": "repack copied ~2.3 GiB of weights into anonymous memory and the phone killed the 4B", "consequence": "Android-only variant frozen; iOS unchanged pending device evidence", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-679" ], "chapter": "phone", "visibility": "public", "status_label": "kept", "aliases": [ "D-system-01" ] }, { "id": "D-025", "date": "2026-10-03", "time": null, "title": "Move v0.2 plaintext out of researcher-readable paths after an exposure; treat the exposed item as seen", "context": "Final cut 2 (cut 1b + exposure fix + citation fix)", "decision": "Move v0.2 plaintext out of researcher-readable paths after an exposure; treat the exposed item as seen", "alternatives": [ "ignore" ], "rationale": "a trace scan printed one held-out item", "consequence": "git history still held the plaintext → later \"burnt\"", "reversible": "partly (history is permanent)", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": "Verification note: Same date fix as the timeline row: the exposure is dated 10-03, not 10-02 22:05.", "fact_ids": [ "F-138" ], "chapter": "ruler", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-15" ] }, { "id": "D-026", "date": "2026-10-03", "time": null, "title": "Declare three health blind sets dev; commission fresh blind sets from writers with no file access", "context": "Several health blind sets declared dev after traces sat in readable paths", "decision": "Declare three health blind sets dev; commission fresh blind sets from writers with no file access", "alternatives": [ "keep them blind" ], "rationale": "traces sat in readable paths and were used as dev", "consequence": "new sets health-fresh1/2/3, chat-v3, meta-fresh1", "reversible": "n/a", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-304" ], "chapter": "builds", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-16" ] }, { "id": "D-027", "date": "2026-10-03", "time": null, "title": "No human review: an independent Opus reviewer agent decides content/wording; caveat \"AI-reviewed, not clinically reviewed\" stays", "context": "\"No human review\": an independent Opus reviewer agent decides content; caveat \"AI-reviewed, not clinically reviewed\"", "decision": "No human review: an independent Opus reviewer agent decides content/wording; caveat \"AI-reviewed, not clinically reviewed\" stays", "alternatives": [ "human/clinical sign-off", "Author or clinician signs off" ], "rationale": "speed; author unavailable for 39 card decisions", "consequence": "reviewer changed 25/31 cards", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1219", "F-834" ], "chapter": "agent-lab", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-17", "D-process-06" ] }, { "id": "D-028", "date": "2026-10-03", "time": null, "title": "Single goal = win the bounty; English only; no more work inside the maintainer's app; build a dedicated measurement app", "context": "No more work inside the app maintainer's app; build \"BOAR Bench\" measurement app", "decision": "Single goal = win the bounty; English only; no more work inside the maintainer's app; build a dedicated measurement app", "alternatives": [ "keep improving the general product" ], "rationale": "focus", "consequence": "PT/ES lines parked; BOAR Bench built", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-075" ], "chapter": "question", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-18" ] }, { "id": "D-029", "date": "2026-10-03", "time": null, "title": "Park the retrieval line after rv2 failed despite more gold in context", "context": "Retrieval v2 killed: gold-in-context 7 → 18 yet quality −0.021", "decision": "Park the retrieval line after rv2 failed despite more gold in context", "alternatives": [ "another retrieval variant" ], "rationale": "gold-sentence-in-context did not translate into judged quality", "consequence": "retrieval reopened later only for the deep tier", "reversible": "yes", "status": "superseded", "status_note": "retrieval reopened later only for the deep tier", "superseded_by": null, "verification_note": null, "fact_ids": [ "F-572" ], "chapter": "deep-tier", "visibility": "public", "status_label": "superseded", "aliases": [ "D-chron-early-19" ] }, { "id": "D-030", "date": "2026-10-03", "time": null, "title": "Single goal = the bounty bar; other quality work after the bounty", "context": "Direction change: single goal = win the bounty; lanes re-tasked", "decision": "Single goal = the bounty bar; other quality work after the bounty", "alternatives": [ "keep broad product work" ], "rationale": "deadline + focus", "consequence": "lanes re-assigned (model tier, health minimum, speed, demo device)", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-076" ], "chapter": "question", "visibility": "public", "status_label": "kept", "aliases": [ "D-system-02" ] }, { "id": "D-031", "date": "2026-10-04", "time": null, "title": "Focus on the 4B default tier; deep tier parked at D1; coordinator runs work with subagents instead of a standing chat team", "context": "Direction: 4B default tier only; deep tier parked; no team chat agents, coordinator + subagents", "decision": "Focus on the 4B default tier; deep tier parked at D1; coordinator runs work with subagents instead of a standing chat team", "alternatives": [ "keep deep tier active" ], "rationale": "token budget; no 12 GB phone", "consequence": "4B levers exhausted by 10-05; deep tier reopened 10-06", "reversible": "yes", "status": "superseded", "status_note": "4B levers exhausted by 10-05; deep tier reopened 10-06", "superseded_by": null, "verification_note": null, "fact_ids": [ "F-704" ], "chapter": "phone", "visibility": "public", "status_label": "superseded", "aliases": [ "D-chron-early-20" ] }, { "id": "D-032", "date": "2026-10-04", "time": null, "title": "Adopt cut 3 (verdict guard + arithmetic repair + snakebite text)", "context": "Cut 3 adopted (cut 2 + verdict guard v3 + arithmetic repair + snakebite text); L1 Δ +0.029 [+0.007, +0.058]", "decision": "Adopt cut 3 (verdict guard + arithmetic repair + snakebite text)", "alternatives": [ "computed-verdict override family (all failed gates)", "verdict guard v4, water card, computed verdict" ], "rationale": "L1 Δ +0.029 [+0.007, +0.058], confident-wrong 15 → 9, unsafe 0", "consequence": "Cut 3 became the first claim build; claim runs 10-05", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1220", "F-1221" ], "chapter": "honest-number", "visibility": "public", "status_label": "kept", "aliases": [ "D-results-01", "D-system-03" ] }, { "id": "D-033", "date": "2026-10-04", "time": null, "title": "Park the web-only knowledge pack for the bounty (public-figure blog pack and a research-forum pack were the alternatives).", "context": "Knowledge-pack sizing: web-only pack parked for the bounty", "decision": "Park the web-only knowledge pack for the bounty (public-figure blog pack and a research-forum pack were the alternatives).", "alternatives": [ "public-figure blog pack, ethresear.ch pack" ], "rationale": "gains come from non-redistributable sources or in-kit misses; best pack +0.007 net", "consequence": "later replaced by the open-corpus Knowledge Pack program (10-12)", "reversible": "yes", "status": "superseded", "status_note": "later replaced by the open-corpus Knowledge Pack program (10-12)", "superseded_by": null, "verification_note": null, "fact_ids": [ "F-084" ], "chapter": "question", "visibility": "public", "status_label": "superseded", "aliases": [ "D-system-04" ] }, { "id": "D-034", "date": "2026-10-05", "time": null, "title": "Fix the KV session save path (KVFIX) and block fragment/figurative instant answers (INSTANT-FRAG) → cut 4", "context": "Cut 4 = cut 3 + KVFIX + INSTANT-FRAG (+ INSTANT-SUBJ)", "decision": "Fix the KV session save path (KVFIX) and block fragment/figurative instant answers (INSTANT-FRAG) → cut 4", "alternatives": [ "skip the route gate; threads; repack" ], "rationale": "A15 smoke: every question re-prefilled the prefixes; \"capital of France\" answered with a Lyon fragment", "consequence": "cut 4 built; KVFIX real gate is the device", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-705" ], "chapter": "phone", "visibility": "public", "status_label": "kept", "aliases": [ "D-system-05" ] }, { "id": "D-035", "date": "2026-10-05", "time": null, "title": "Bench APK with INTERNET permission removed (no-network by construction)", "context": "No-network measurement APK (INTERNET permission removed by construction)", "decision": "Bench APK with INTERNET permission removed (no-network by construction)", "alternatives": [ "rely on airplane mode only" ], "rationale": "prove offline operation statically", "consequence": "audit verdict NO-NETWORK-BY-CONSTRUCTION; later phone runs used it", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-706" ], "chapter": "phone", "visibility": "public", "status_label": "kept", "aliases": [ "D-system-06" ] }, { "id": "D-036", "date": "2026-10-06", "time": null, "title": "Headline = floor-adjusted ratio + win/tie/loss (+ share of items ≥ 3/5), raw second; own bar = floor-adj ≥ 0.50 on V-para", "context": "Honest-metric rewrite: headline becomes floor-adjusted ratio + win/tie/loss, raw second", "decision": "Headline = floor-adjusted ratio + win/tie/loss (+ share of items ≥ 3/5), raw second; own bar = floor-adj ≥ 0.50 on V-para", "alternatives": [ "keep raw headline", "Raw headline" ], "rationale": "raw 0.50 is ~61 % free because of the rubric floor", "consequence": "board and claim draft rewritten 10-06 22:40", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": "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)", "fact_ids": [ "F-1146", "F-1226", "F-1227" ], "chapter": "honest-number", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-21", "D-method-09", "D-process-08" ] }, { "id": "D-037", "date": "2026-10-06", "time": null, "title": "Treat v0.2 as burnt (dev); a final number needs v0.3 written outside the Claude family", "context": "Retrieval scorecard: coverage is the largest failure bucket (49 %); recall@5 0.333", "decision": "Treat v0.2 as burnt (dev); a final number needs v0.3 written outside the Claude family", "alternatives": [ "rewrite git history", "keep using v0.2 as held-out" ], "rationale": "plaintext in private git history + 4 runs + a targeted fix", "consequence": "v0.3 spec written 10-07; items not written by 10-10", "reversible": "partly", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": "citation fix: D5 is l.57", "fact_ids": [ "F-1248", "F-1249" ], "chapter": "handoff", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-22", "D-method-10" ] }, { "id": "D-038", "date": "2026-10-06", "time": null, "title": "Freeze the claim build at cut 6; later only routing/safety fixes that leave v0.1 answers unchanged", "context": "Claim/device build frozen at cut 6; later only routing/safety fixes", "decision": "Freeze the claim build at cut 6; later only routing/safety fixes that leave v0.1 answers unchanged", "alternatives": [ "keep shipping answer-changing arms" ], "rationale": "stable claim evidence", "consequence": "cut 7 later added the fact table (0 fires on held-out sets)", "reversible": "yes", "status": "superseded", "status_note": "Relaxed on 2026-10-07: cut 7 added FACTS r6, which changed 8 v0.1 answers.", "superseded_by": "D-046", "verification_note": "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)", "fact_ids": [ "F-331", "F-723" ], "chapter": "builds", "visibility": "public", "status_label": "superseded by D-046", "aliases": [ "D-chron-early-23", "D-system-09" ] }, { "id": "D-039", "date": "2026-10-06", "time": null, "title": "Stop answer-identical 4B arms; reopen the deep tier as the primary quality lever", "context": "Retrieval scorecard: coverage is the largest failure bucket (49 %); recall@5 0.333", "decision": "Stop answer-identical 4B arms; reopen the deep tier as the primary quality lever", "alternatives": [ "continue 4B polishing" ], "rationale": "15 arms, 0 ratio movement", "consequence": "deep-tier line ran 10-07 → 10-10", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-576" ], "chapter": "deep-tier", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-24" ] }, { "id": "D-040", "date": "2026-10-06", "time": "21:25", "title": "Cheap judge for triage (Sonnet-low, replaced by Haiku 5.5-low on 10-07); Opus only for confirmation and claim numbers", "context": "Cheaper judge triage adopted (Sonnet-low, later Haiku)", "decision": "Cheap judge for triage (Sonnet-low, replaced by Haiku 5.5-low on 10-07); Opus only for confirmation and claim numbers", "alternatives": [ "Opus for everything" ], "rationale": "judge is the largest cost", "consequence": "judge calls fell from 12k–20k calls / 6 days to ≈ 1.5k / 2 days", "reversible": "yes", "status": "superseded", "status_note": "Sonnet-low was replaced by Claude Haiku 5.5-low as the L1 judge on 2026-10-07 (HAIKUJUDGE).", "superseded_by": "D-049", "verification_note": null, "fact_ids": [ "F-147", "F-157" ], "chapter": "ruler", "visibility": "public", "status_label": "superseded by D-049", "aliases": [ "D-chron-early-25", "D-method-11" ] }, { "id": "D-041", "date": "2026-10-06", "time": null, "title": "Declare v0.2 burnt → dev after its plaintext was found in git history", "context": "v0.2 ruled burnt (plaintext in private bench repo history) → dev", "decision": "Declare v0.2 burnt → dev after its plaintext was found in git history", "alternatives": [ "Keep calling it held-out" ], "rationale": "Honesty: builders could have seen it", "consequence": "An unexposed held-out number needs a v0.3 written outside the Claude family (0 items as of 10-17; open front OF-3)", "reversible": "no", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": "Verification note: Public v0.2 row is SET-STATUS:14 (not :17).", "fact_ids": [ "F-1250" ], "chapter": "handoff", "visibility": "public", "status_label": "kept", "aliases": [ "D-process-07" ] }, { "id": "D-042", "date": "2026-10-06", "time": null, "title": "Keep FACTS r6 as head despite FAIL by frozen rule (recall 70 % < 75 %)", "context": "FACTS r6 on blind set 4: recall 70 % < 75 % → FAIL; kept as head by decision", "decision": "Keep FACTS r6 as head despite FAIL by frozen rule (recall 70 % < 75 %)", "alternatives": [ "close at r5; build r7 (forbidden by criterion)" ], "rationale": "r6 dominates r5 on every frozen measure; precision 96.6 %, FP 0/100 adversarial", "consequence": "Shipped in cut 7; stated as \"FAIL stands\"", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-332" ], "chapter": "builds", "visibility": "public", "status_label": "kept", "aliases": [ "D-results-03" ] }, { "id": "D-043", "date": "2026-10-06", "time": null, "title": "Lead every public number with floor-adjusted + win-rate, not raw", "context": "Honest-metric rewrite: headline becomes floor-adjusted ratio + win/tie/loss, raw second", "decision": "Lead every public number with floor-adjusted + win-rate, not raw", "alternatives": [ "keep raw as headline (pre-registered formal rule)" ], "rationale": "raw 0.50 is a rubric-floor artefact (always-decline ≈ 0.31 raw)", "consequence": null, "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": "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", "fact_ids": [ "F-277" ], "chapter": "floor-trap", "visibility": "public", "status_label": "kept", "aliases": [ "D-results-04" ] }, { "id": "D-044", "date": "2026-10-06", "time": null, "title": "Adopt CUT5-LITE (KV prefix for no-source prompts + route-gate skip), reject CUT5-SPEED", "context": "Cut 5 = cut 4 + no-source KV prefix + route-gate skip (CUT5-LITE adopted; CUT5-SPEED rejected on desktop gate)", "decision": "Adopt CUT5-LITE (KV prefix for no-source prompts + route-gate skip), reject CUT5-SPEED", "alternatives": [ "ship all speed arms" ], "rationale": "CUT5-SPEED failed its desktop speed gate (1.071× vs ≤ 0.8); lite was identity-safe", "consequence": "cut 5 = device-day candidate", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-722" ], "chapter": "phone", "visibility": "public", "status_label": "kept", "aliases": [ "D-system-07" ] }, { "id": "D-045", "date": "2026-10-06", "time": null, "title": "Adopt DETMAX → cut 6", "context": "Cut 6 = cut 5 + DETMAX (detector scores window and current turn)", "decision": "Adopt DETMAX → cut 6", "alternatives": [ "no change; lexical follow-up fix (HFU, failed)" ], "rationale": "chit-chat before an emergency lost 7/240 routes; DETMAX 0/240", "consequence": null, "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": "citation fix (l.46 is blank)", "fact_ids": [ "F-333" ], "chapter": "builds", "visibility": "public", "status_label": "kept", "aliases": [ "D-system-08" ] }, { "id": "D-046", "date": "2026-10-07", "time": "09:23", "title": "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\"", "context": "Cut 7 = cut 6 + FACTS r6; v0.3 to be written outside the Claude family; \"no device exists\"", "decision": "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\"", "alternatives": [ "wait for more recall", "re-run full claim runs (burns held-out again)", "no FACTS; r7" ], "rationale": "deterministic facts lever; no phone", "consequence": "cut 7 frozen; fine-tune data failed review same day", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1230", "F-353", "F-745" ], "chapter": "phone", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-26", "D-results-02", "D-system-10" ] }, { "id": "D-047", "date": "2026-10-07", "time": null, "title": "v0.3 written by the author with a non-Claude web model; hashes only before scoring", "context": "Bench v0.3 spec accepted: claim set to be written outside the Claude family by the author with a non-Claude web-UI model", "decision": "v0.3 written by the author with a non-Claude web model; hashes only before scoring", "alternatives": [ "Claude-written items", "Claude subagent writers (as v0.2)" ], "rationale": "self-preference risk; burnt v0.2", "consequence": "spec ready; items not written by 10-10", "reversible": "yes", "status": "superseded", "status_note": "On 2026-10-18 v0.3 was dropped before closure (D5); the v0.4 ruler set replaced it.", "superseded_by": "D-123", "verification_note": null, "fact_ids": [ "F-1148", "F-1251" ], "chapter": "measurer", "visibility": "public", "status_label": "superseded by D-123", "aliases": [ "D-chron-early-27", "D-method-15" ] }, { "id": "D-048", "date": "2026-10-07", "time": "14:26", "title": "Autonomous mandate: keep running rounds without asking; stop only for irreversible/expensive items", "context": "JUDGE2: local open-weights judge (Gemma 4 26B) keeps the ranking but scores +0.11 above Opus", "decision": "Autonomous mandate: keep running rounds without asking; stop only for irreversible/expensive items", "alternatives": [ "ask per round" ], "rationale": "throughput", "consequence": "deep-tier rounds ran unattended", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-588" ], "chapter": "deep-tier", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-28" ] }, { "id": "D-049", "date": "2026-10-07", "time": null, "title": "Haiku 5.5-low replaces Sonnet-low as L1 judge", "context": "Haiku 5.5-low becomes the L1 triage judge; judge sharding N = 4; \"no device exists\" ruling", "decision": "Haiku 5.5-low replaces Sonnet-low as L1 judge", "alternatives": [ "Sonnet-low; Haiku medium" ], "rationale": "same winner 54/57, 3.5 s/call, ~14× cheaper", "consequence": "adopted; known to compress gaps", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1149" ], "chapter": "measurer", "visibility": "public", "status_label": "kept", "aliases": [ "D-method-12" ] }, { "id": "D-050", "date": "2026-10-07", "time": null, "title": "Local open-weights judge (Gemma 4 26B) allowed for triage/ranking only, never for the number", "context": "JUDGE2: local open-weights judge (Gemma 4 26B) keeps the ranking but scores +0.11 above Opus", "decision": "Local open-weights judge (Gemma 4 26B) allowed for triage/ranking only, never for the number", "alternatives": [ "adopt as main judge" ], "rationale": "keeps ranking, +0.11 lenient, κ 0.32-0.55", "consequence": "not used for headline numbers", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-283" ], "chapter": "floor-trap", "visibility": "public", "status_label": "kept", "aliases": [ "D-method-13" ] }, { "id": "D-051", "date": "2026-10-07", "time": null, "title": "Shard the judge: N = 4 default for Haiku L1 and for Opus L2/claim", "context": "JUDGE2: local open-weights judge (Gemma 4 26B) keeps the ranking but scores +0.11 above Opus", "decision": "Shard the judge: N = 4 default for Haiku L1 and for Opus L2/claim", "alternatives": [ "sequential; N = 8" ], "rationale": "×4.5 wall gain, 0 rate limits, same winners", "consequence": "in force; a stopped shard is resumed, never skipped", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-159" ], "chapter": "ruler", "visibility": "public", "status_label": "kept", "aliases": [ "D-method-14" ] }, { "id": "D-052", "date": "2026-10-07", "time": null, "title": "Haiku 5.5-low as L1 triage judge, Opus only to confirm and for claim numbers; judge sharded N = 4", "context": "Haiku 5.5-low becomes the L1 triage judge; judge sharding N = 4; \"no device exists\" ruling", "decision": "Haiku 5.5-low as L1 triage judge, Opus only to confirm and for claim numbers; judge sharded N = 4", "alternatives": [ "Opus everywhere; Sonnet-low triage", "Sonnet-low; local Gemma (ranking only, +0.11 lenient)" ], "rationale": "≈ 14× cheaper than the previous triage judge; same winner as Opus 54/57", "consequence": "Haiku compresses gaps 15-30 %; near-miss escalation rules added", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1150", "F-1151" ], "chapter": "measurer", "visibility": "public", "status_label": "kept", "aliases": [ "D-process-09", "D-results-05" ] }, { "id": "D-053", "date": "2026-10-07", "time": null, "title": "\"No device exists\": device tasks dormant", "context": "Haiku 5.5-low becomes the L1 triage judge; judge sharding N = 4; \"no device exists\" ruling", "decision": "\"No device exists\": device tasks dormant", "alternatives": [ "Keep proposing device work" ], "rationale": "No 12 GB phone available", "consequence": "Later reversed in practice when a 10.7 GiB phone ran cut 7 on 10-14", "reversible": "yes", "status": "reversed", "status_note": "A 12 GB-class phone (10.7 GiB usable) ran cut 7 on 2026-10-14.", "superseded_by": null, "verification_note": null, "fact_ids": [ "F-733" ], "chapter": "phone", "visibility": "public", "status_label": "reversed", "aliases": [ "D-process-10" ] }, { "id": "D-054", "date": "2026-10-08", "time": null, "title": "SLICE20 triage before any full deep/held-out arm", "context": "SLICE20: every deep-tier / held-out-style arm starts on a 20-item slice", "decision": "SLICE20 triage before any full deep/held-out arm", "alternatives": [ "full sets always" ], "rationale": "~60 % generation saving on NO arms", "consequence": "later replaced by stratified random slices (10-10)", "reversible": "yes", "status": "superseded", "status_note": "later replaced by stratified random slices (10-10)", "superseded_by": null, "verification_note": null, "fact_ids": [ "F-169" ], "chapter": "ruler", "visibility": "public", "status_label": "superseded", "aliases": [ "D-chron-early-29" ] }, { "id": "D-055", "date": "2026-10-09", "time": null, "title": "V-para status: held-out for the 4B claim runs; dev-style for everything measured after 10-06", "context": "V-para status ruling: held-out for the 4B claim runs, dev-style for everything measured after 2026-10-06", "decision": "V-para status: held-out for the 4B claim runs; dev-style for everything measured after 10-06", "alternatives": [ "keep calling it blind", "keep calling it held-out" ], "rationale": "exposure ledger showed conflicting status labels", "consequence": "every deep-tier number on V-para is labelled dev-style", "reversible": "no (status)", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1232", "F-1233" ], "chapter": "honest-number", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-30", "D-method-16" ] }, { "id": "D-056", "date": "2026-10-09", "time": null, "title": "Recommend thinking budget 512 for the deep tier", "context": "DT512: thinking budget 512 = 1024 quality at 0.55× time → recommended config", "decision": "Recommend thinking budget 512 for the deep tier", "alternatives": [ "1024; 256", "1024 (slower), 256 (−0.028, CHEAPER-LOSS), Gemma + thinking (−0.028)", "256 (loses), 1024, 3072, Gemma 26B" ], "rationale": "CHEAPER-PARITY with 1024 at 0.55× time", "consequence": "becomes the \"recommended deep configuration\"", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-598", "F-599", "F-748" ], "chapter": "deep-tier", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-31", "D-results-06", "D-system-11" ] }, { "id": "D-057", "date": "2026-10-10", "time": null, "title": "Near-miss rule: a Haiku Δ within 0.01 of a threshold triggers Opus on the same ids", "context": "Near-miss rule after a Haiku triage stopped an arm Opus later rated +0.048", "decision": "Near-miss rule: a Haiku Δ within 0.01 of a threshold triggers Opus on the same ids", "alternatives": [ "trust Haiku", "trust Haiku always" ], "rationale": "Haiku compressed a +0.048 Opus gap to +0.027", "consequence": "applied to THINK4B", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1158", "F-613" ], "chapter": "measurer", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-early-32", "D-method-17" ] }, { "id": "D-058", "date": "2026-10-10", "time": null, "title": "Seed-noise reading rule (V-para one-seed deep Δ < ~0.03 is noise; later raised to 0.05)", "context": "SEEDVAR: H over 3 seeds mean floor-adj 0.432 (SD 0.025); one-seed deep Δ < 0.03 = seed noise", "decision": "Seed-noise reading rule (V-para one-seed deep Δ < ~0.03 is noise; later raised to 0.05)", "alternatives": [ "single-seed verdicts", "single-seed decisions", "treat one-seed rows as decisive" ], "rationale": "SEEDVAR three-seed spread", "consequence": "in CLAUDE.md and RULE-CHANGES", "reversible": "yes", "status": "superseded", "status_note": "Threshold raised from 0.03 to 0.05 on 2026-10-16 after the outside critique (SEEDVAR showed a −0.039 gap between seeds).", "superseded_by": "D-113", "verification_note": null, "fact_ids": [ "F-1159", "F-1160", "F-1162" ], "chapter": "measurer", "visibility": "public", "status_label": "superseded by D-113", "aliases": [ "D-chron-early-33", "D-method-18", "D-results-08" ] }, { "id": "D-059", "date": "2026-10-10", "time": "21:54", "title": "Research on the Mac mini only until the bar is met; no APK/device/claim-package work", "context": "Author's direction: research on the Mac mini only until the bar is met; device and claim work after", "decision": "Research on the Mac mini only until the bar is met; no APK/device/claim-package work", "alternatives": [ "claim-path first (review's recommendation)", "A deep-tier arms; B v0.3 outside Claude; C bigger model" ], "rationale": "author's direction after the process review", "consequence": "handoff rewritten for research scope", "reversible": "yes", "status": "superseded", "status_note": "The closing sprint from 2026-10-14 brought app and release work back (upstream PR, public repositories).", "superseded_by": null, "verification_note": null, "fact_ids": [ "F-763", "F-854" ], "chapter": "phone", "visibility": "public", "status_label": "superseded", "aliases": [ "D-chron-early-34", "D-chron-late-02" ] }, { "id": "D-060", "date": "2026-10-10", "time": "21:05", "title": "Proceed on stated defaults for the claim path (debug-signed APK, raw formal + floor-adj + win-rate, no v0.3, no device relay)", "context": "Decision note with defaults sent to the author (claim vehicle, bar scale \"raw formal + floor-adj + win-rate\", no v0.3, no device relay)", "decision": "Proceed on stated defaults for the claim path (debug-signed APK, raw formal + floor-adj + win-rate, no v0.3, no device relay)", "alternatives": [ "ask and wait" ], "rationale": "unblock work without waiting on the author", "consequence": "superseded 45 min later by the STOP", "reversible": "yes", "status": "superseded", "status_note": "superseded 45 min later by the STOP", "superseded_by": null, "verification_note": null, "fact_ids": [ "F-758" ], "chapter": "phone", "visibility": "public", "status_label": "superseded", "aliases": [ "D-chron-late-01" ] }, { "id": "D-061", "date": "2026-10-10", "time": "22:09", "title": "Every deciding run needs a frozen criterion with a Power block; slices = random20 files", "context": "Criterion template with mandatory Power block; random-20 slices; bar table frozen", "decision": "Every deciding run needs a frozen criterion with a Power block; slices = random20 files", "alternatives": [ "signal20 slices (retired)", "voluntary", "Optional" ], "rationale": "stop under-powered closures", "consequence": "arm.sh refuses launch without the block", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-177", "F-180", "F-195" ], "chapter": "ruler", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-03", "D-method-21", "D-process-12" ] }, { "id": "D-062", "date": "2026-10-10", "time": null, "title": "Decide direction on v0.2 (seed-stable), confirm on V-para with 2 seeds; adopt a lever only after two-seed V-para confirmation", "context": "Criterion template with mandatory Power block; random-20 slices; bar table frozen", "decision": "Decide direction on v0.2 (seed-stable), confirm on V-para with 2 seeds; adopt a lever only after two-seed V-para confirmation", "alternatives": [ "decide on V-para one seed" ], "rationale": "v0.2 per-item seed r 0.90 vs V-para 0.5-0.7", "consequence": "in force", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-176" ], "chapter": "ruler", "visibility": "public", "status_label": "kept", "aliases": [ "D-method-19" ] }, { "id": "D-063", "date": "2026-10-10", "time": null, "title": "Fail-closed pipeline: errored traces/missing ids stop the run or score at the floor", "context": "Fail-closed judge pipeline: errored or missing items now stop the run or score at the floor instead of vanishing", "decision": "Fail-closed pipeline: errored traces/missing ids stop the run or score at the floor", "alternatives": [ "drop silently", "Manual checks" ], "rationale": "the old judge silently dropped errored traces, flattering a failing device", "consequence": "flags `--strict`, `--floor-errored`, `--floor-missing`", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-179", "F-752" ], "chapter": "phone", "visibility": "public", "status_label": "kept", "aliases": [ "D-method-20", "D-process-11" ] }, { "id": "D-064", "date": "2026-10-10", "time": null, "title": "Random stratified 20-id slices (seed 7) replace the \"20 hardest items\" slice", "context": "Power block mandatory in every criterion; random stratified 20-id slices replace the \"20 hardest\" slice", "decision": "Random stratified 20-id slices (seed 7) replace the \"20 hardest items\" slice", "alternatives": [ "keep signal20", "Hardest-20 slices" ], "rationale": "hardest-20 inflated deltas (+0.091 slice vs +0.026 full)", "consequence": "in force; slice only screens for futility", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1161", "F-178" ], "chapter": "ruler", "visibility": "public", "status_label": "kept", "aliases": [ "D-method-22", "D-process-13" ] }, { "id": "D-065", "date": "2026-10-10", "time": null, "title": "Win-rate reported, never gated (later: proposal to drop the win-rate conjunct because it is saturated at 0)", "context": "Author's direction: research on the Mac mini only until the bar is met; device and claim work after", "decision": "Win-rate reported, never gated (later: proposal to drop the win-rate conjunct because it is saturated at 0)", "alternatives": [ "gate on win-rate > 0" ], "rationale": "every phone system scores 0 wins vs a 4.9/5 reference", "consequence": null, "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": "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", "fact_ids": [ "F-753" ], "chapter": "phone", "visibility": "public", "status_label": "kept", "aliases": [ "D-method-28" ] }, { "id": "D-067", "date": "2026-10-10", "time": null, "title": "Declare the deep-tier line closed on desktop and the 4B side closed", "context": "Desktop queue empty: 4B side and deep-tier line both closed", "decision": "Declare the deep-tier line closed on desktop and the 4B side closed", "alternatives": [ "keep sweeping configs" ], "rationale": "model, parity, thinking, budget, persona/context, routing all measured", "consequence": "Research moved to Knowledge Packs (10-12)", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-614" ], "chapter": "deep-tier", "visibility": "public", "status_label": "kept", "aliases": [ "D-results-07" ] }, { "id": "D-068", "date": "2026-10-10", "time": null, "title": "Keep retrieval on the deep path", "context": "DEEPRERANK, MP4B, DEEPITER, DEEPCOV-FB closed (retrieval-side levers on the deep tier)", "decision": "Keep retrieval on the deep path", "alternatives": [ "drop packs for speed" ], "rationale": "no-pack −0.045 (−0.083…−0.011)", "consequence": "kept", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-997" ], "chapter": "packs", "visibility": "public", "status_label": "kept", "aliases": [ "D-system-12" ] }, { "id": "D-069", "date": "2026-10-11", "time": "03:4x", "title": "Run research as an unattended self-paced loop (one experiment per tick, META every 8 ticks)", "context": "LOOP-PROGRAM written: unattended self-paced loop, one experiment per tick", "decision": "Run research as an unattended self-paced loop (one experiment per tick, META every 8 ticks)", "alternatives": [ "session-by-session manual runs", "Author-driven sessions" ], "rationale": "use idle machine time; keep method improving", "consequence": "133 ticks over 10-11→10-14", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-857", "F-858" ], "chapter": "agent-lab", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-04", "D-process-15" ] }, { "id": "D-070", "date": "2026-10-11", "time": "12:10", "title": "Measure the loss map before more levers (LOSSMAP first)", "context": "Loop tick 1: LOSSMAP measures where the deep tier loses points", "decision": "Measure the loss map before more levers (LOSSMAP first)", "alternatives": [ "keep running backlog order" ], "rationale": "direct levers at the biggest loss bucket", "consequence": "re-ranked backlog; absent knowledge = 0.59 / 0.55 of loss", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-860" ], "chapter": "agent-lab", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-05" ] }, { "id": "D-072", "date": "2026-10-11", "time": null, "title": "Common-mode CIs for deep-tier rows: CI = Δ ± 1.96·sqrt(SE_item² + 2σ_run²)", "context": "JUDGEVAR: re-judging unchanged answers gives σ_judge ≈ 0.003; common-mode CIs adopted for deep-tier rows", "decision": "Common-mode CIs for deep-tier rows: CI = Δ ± 1.96·sqrt(SE_item² + 2σ_run²)", "alternatives": [ "item bootstrap only", "plain item bootstrap" ], "rationale": "item bootstrap ignores run-level seed noise", "consequence": "deep CIs ~1.2-1.3× wider; readings unchanged", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1176", "F-1236" ], "chapter": "measurer", "visibility": "public", "status_label": "kept", "aliases": [ "D-method-23", "D-results-10" ] }, { "id": "D-073", "date": "2026-10-11", "time": null, "title": "Negative results relabelled \"not detected at MDE x (n, seeds)\" when under-powered", "context": "CLOSED.md registry built (36 rows); honesty relabel \"not detected at MDE x\" for under-powered negatives", "decision": "Negative results relabelled \"not detected at MDE x (n, seeds)\" when under-powered", "alternatives": [ "keep \"closed\"", "Plain \"closed\"" ], "rationale": "\"closed\" overstated weak negatives", "consequence": "CLOSED.md rows carry MDE and reopen condition", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": "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)", "fact_ids": [ "F-1177", "F-1178", "F-198" ], "chapter": "measurer", "visibility": "public", "status_label": "kept", "aliases": [ "D-method-24", "D-process-17", "D-results-09" ] }, { "id": "D-074", "date": "2026-10-11", "time": null, "title": "Verdict labels LEVER / SMALL / NO / \"not decided at MDE\"; adopt only after two-seed confirmation on V-para", "context": "Verdict labels LEVER / SMALL / NO / \"not detected at MDE\"; honesty relabel of under-powered closures", "decision": "Verdict labels LEVER / SMALL / NO / \"not decided at MDE\"; adopt only after two-seed confirmation on V-para", "alternatives": [ "Point-estimate promotion" ], "rationale": "Stop over-reading noise", "consequence": "Adopted levers after 10-08: 0 on the bounty set", "reversible": "no", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-087" ], "chapter": "question", "visibility": "public", "status_label": "kept", "aliases": [ "D-process-16" ] }, { "id": "D-075", "date": "2026-10-12", "time": "07:55", "title": "Close QUANTCEIL as \"not detected at MDE 0.03\" after V-para two seeds", "context": "QUANTCEIL: 3-bit quant of deep model +0.004 (not detected); cycle adds 0 V-para points per 1,804 judge calls", "decision": "Close QUANTCEIL as \"not detected at MDE 0.03\" after V-para two seeds", "alternatives": [ "adopt on v0.2 wrong-bucket repair" ], "rationale": "seeds split +0.030 / −0.022", "consequence": "model-knowledge levers on this hardware declared exhausted", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-207" ], "chapter": "ruler", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-07" ] }, { "id": "D-076", "date": "2026-10-12", "time": "14:14", "title": "Withdraw \"facts are not in Wikipedia\"; adopt the peer research's task chain", "context": "Correction: \"facts are not in Wikipedia\" withdrawn (scope was the top-100k pack only)", "decision": "Withdraw \"facts are not in Wikipedia\"; adopt the peer research's task chain", "alternatives": [ "keep the closure", "build a new corpus right away" ], "rationale": "peer review showed WIKISECTION covered only one pack and part of the items", "consequence": "WIKIFULL stages run; result 9/34 (a lower bound)", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": "Verification note: Outcome should read \"9/34 verified (lower bound, MARGINAL)\", not \"9/34 (a lower bound)\".", "fact_ids": [ "F-1003", "F-1011" ], "chapter": "packs", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-08", "D-results-11" ] }, { "id": "D-077", "date": "2026-10-12", "time": "14:14", "title": "Phone-time wording: deep + think 512 ≈ 1.9–2.8 min on a Pixel 8 Pro CPU; competitor percentages never quoted as a gap", "context": "SYNTHDEEP, WIKISECTION, DEEPHEDGE closed at 0-call gates", "decision": "Phone-time wording: deep + think 512 ≈ 1.9–2.8 min on a Pixel 8 Pro CPU; competitor percentages never quoted as a gap", "alternatives": [ "the earlier 3.5–5 min figure" ], "rationale": "corrected premises 2 and 5", "consequence": "applied in all later texts", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-768" ], "chapter": "phone", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-09" ] }, { "id": "D-078", "date": "2026-10-12", "time": "20:06", "title": "Pivot to Knowledge Packs; primary metric = in-pack conditional ratio", "context": "Author's direction: Knowledge Packs; research question becomes conditional quality with the matching pack installed", "decision": "Pivot to Knowledge Packs; primary metric = in-pack conditional ratio", "alternatives": [ "stay on the bar", "keep deep-tier sweeps; non-Wikipedia corpus" ], "rationale": "the product is a small model + user-selected packs; the unconditional bar was not moving", "consequence": "P1–P4, pack workflow, pack stack", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1004", "F-1012" ], "chapter": "packs", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-10", "D-results-12" ] }, { "id": "D-079", "date": "2026-10-12", "time": "22:32", "title": "Build a pack-authoring workflow; agents may write the 20 public self-tests per pack; held-out claim items stay non-Claude", "context": "Author: build a pack-authoring workflow (\"we will be the first author with publications\")", "decision": "Build a pack-authoring workflow; agents may write the 20 public self-tests per pack; held-out claim items stay non-Claude", "alternatives": [ "only use existing packs" ], "rationale": "\"we will be the first author with publications\"", "consequence": "11 packs published 10-13", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1015" ], "chapter": "packs", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-11" ] }, { "id": "D-080", "date": "2026-10-12", "time": "22:46", "title": "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, no", "context": "Pack catalogue + authoring workflow (licence policy PD/CC0/CC BY/CC BY-SA only)", "decision": "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", "alternatives": [ "format v3 now; NC corpora (wikiHow, WHO, OpenStax)", "Kiwix 2026 ZIMs (fresher)" ], "rationale": "zero format risk; licence compliance", "consequence": "applied to all published packs", "reversible": "partly (published data stays public)", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": "citation fix: D-K5 is l.12 (l.13 is D-K6)", "fact_ids": [ "F-1017", "F-368" ], "chapter": "packs", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-12", "D-system-14" ] }, { "id": "D-081", "date": "2026-10-12", "time": "23:22", "title": "Autonomy rule: never ask for approval; paid API only when no free equivalent, capped and logged", "context": "Autonomy rule from the author: never ask for approval; paid API only when no free equivalent", "decision": "Autonomy rule: never ask for approval; paid API only when no free equivalent, capped and logged", "alternatives": [ "per-step approval" ], "rationale": "author said Claude usage is not the constraint", "consequence": "loop ran without approval asks", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-870" ], "chapter": "agent-lab", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-13" ] }, { "id": "D-082", "date": "2026-10-12", "time": null, "title": "Knowledge packs = format 2 SQLite, per-row attribution, four artefacts per pack (card, self-test, coverage probe, crowding probe)", "context": "PACKCOND P1–P4, PACKDENSE, HEALTHQ, SECTIONHIT: in-pack loss is retrieval, not grounding", "decision": "Knowledge packs = format 2 SQLite, per-row attribution, four artefacts per pack (card, self-test, coverage probe, crowding probe)", "alternatives": [ "format v3 first" ], "rationale": "zero format risk; licence compliance; measured crowding failure", "consequence": "11 packs built and published 10-13", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1005" ], "chapter": "packs", "visibility": "public", "status_label": "kept", "aliases": [ "D-system-13" ] }, { "id": "D-083", "date": "2026-10-12", "time": null, "title": "\"Default-eligible\" only if a pack holds ≤ 0.40 of kept slots and does not lower kit gold; otherwise user-selected only", "context": "Author: build a pack-authoring workflow (\"we will be the first author with publications\")", "decision": "\"Default-eligible\" only if a pack holds ≤ 0.40 of kept slots and does not lower kit gold; otherwise user-selected only", "alternatives": [ "install everything" ], "rationale": "crowding measured", "consequence": null, "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": "citation fix: the rule is l.50, statuses in the table l.35-45 (5 default-eligible, 6 user-selected)", "fact_ids": [ "F-1006" ], "chapter": "packs", "visibility": "public", "status_label": "kept", "aliases": [ "D-system-15" ] }, { "id": "D-084", "date": "2026-10-13", "time": "00:01", "title": "Publish packs that pass the four-artefact check (card, self-test, coverage, crowding) on HF under the author's standing OK", "context": "Pack-stack technical report (crowding, routing, keep, reader)", "decision": "Publish packs that pass the four-artefact check (card, self-test, coverage, crowding) on HF under the author's standing OK", "alternatives": [ "hold until measured further" ], "rationale": "public reuse; licence clarity per pack", "consequence": "7 + 2 + 2 packs published", "reversible": "no (public)", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1036" ], "chapter": "packs", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-14" ] }, { "id": "D-085", "date": "2026-10-13", "time": "10:20", "title": "No cap on Haiku judge calls (Rule 25)", "context": null, "decision": "No cap on Haiku judge calls (Rule 25)", "alternatives": [ "weekly Haiku 400 cap" ], "rationale": "Haiku is cheap triage; idle caps cost machine time", "consequence": "720-call scoring passes became routine", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-876" ], "chapter": "agent-lab", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-15" ] }, { "id": "D-086", "date": "2026-10-13", "time": "13:4x", "title": "Run an oracle/ceiling cell (ORACLE-ST) before more pre-processing levers", "context": "A side session re-judged existing traces only: PACKS-L2, ORACLE-ST, RIVALSET, red-team of the claim, pipeline and ledger audits", "decision": "Run an oracle/ceiling cell (ORACLE-ST) before more pre-processing levers", "alternatives": [ "keep adding retrieval tricks" ], "rationale": "decide whether reader or retrieval is the wall", "consequence": "ceiling +0.095 found; became a standing rule (LOOP v2 D8)", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1045" ], "chapter": "packs", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-16" ] }, { "id": "D-087", "date": "2026-10-13", "time": "15:44", "title": "Routing among packs closed at the product level; keep rule + quota only an optional setting", "context": "REPORT-PACKS drafted and independently reviewed (39 edits)", "decision": "Routing among packs closed at the product level; keep rule + quota only an optional setting", "alternatives": [ "top-1 router, 4B zero-shot classifier" ], "rationale": "crowding cost small (−0.016 Haiku / −0.019 Opus); routers 0.25–0.61 accuracy", "consequence": "product reading: ship small user-selected packs, no automatic routing", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1046" ], "chapter": "packs", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-17" ] }, { "id": "D-089", "date": "2026-10-13", "time": "22:08", "title": "Claim draft v3.9 adopts the red-team findings (v0.2 \"formerly held-out\", V-para writer disclosed, verdict reuse disclosed)", "context": "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", "decision": "Claim draft v3.9 adopts the red-team findings (v0.2 \"formerly held-out\", V-para writer disclosed, verdict reuse disclosed)", "alternatives": [ "keep v3.8" ], "rationale": "red-team found no inflated number but 9 overclaiming sentences", "consequence": "41 edits applied", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-216" ], "chapter": "ruler", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-19" ] }, { "id": "D-090", "date": "2026-10-13", "time": null, "title": "Never idle while 0-paid-call work exists; no Haiku cap", "context": "Author challenges 9+ idle ticks; loop un-idled; rules 24 (never idle while 0-call work exists) and 25 (no Haiku cap)", "decision": "Never idle while 0-paid-call work exists; no Haiku cap", "alternatives": [ "Keep caps" ], "rationale": "Author objected after 9 idle ticks (\"why idle when I asked to never stop?\", paraphrase)", "consequence": "10-14 log: 0 idle ticks; later critic warns \"idle lane = failure\" rewards trivial probes", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1185" ], "chapter": "measurer", "visibility": "public", "status_label": "kept", "aliases": [ "D-process-18" ] }, { "id": "D-091", "date": "2026-10-14", "time": "02:53", "title": "Run doc2query PoC overnight on the MacBook instead of renting a GPU", "context": "MICROD2Q (doc2query) fails; doc2query closed", "decision": "Run doc2query PoC overnight on the MacBook instead of renting a GPU", "alternatives": [ "rented GPU" ], "rationale": "cost ≈ US$0.3–1 vs free local hours", "consequence": "4.77 h, FAIL; later cited as an avoidable loss (speed gate overridden by an addendum)", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-921" ], "chapter": "agent-lab", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-20" ] }, { "id": "D-092", "date": "2026-10-14", "time": "15:38", "title": "HEADPASS-X2b = the pack line's retrieval recommendation for format-v3 packs; X1 for format-2 packs", "context": "HEADPASS-X2b lever (+0.066 on 180 dev self-tests)", "decision": "HEADPASS-X2b = the pack line's retrieval recommendation for format-v3 packs; X1 for format-2 packs", "alternatives": [ "reranker X3; unbudgeted X2", "X1 only (+0.033); X3 reranker (FAIL reach, 21 s/item)" ], "rationale": "Opus +0.066 (+0.043…+0.091), no harm on kit", "consequence": "X3 failed at step 0; X2b staged for after 10-20 (PR-2)", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1059", "F-1070" ], "chapter": "packs", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-21", "D-results-13" ] }, { "id": "D-093", "date": "2026-10-14", "time": null, "title": "SFT of the 4B closed unless a teacher reaches ≥ 50 % faithfulness on distractor-rich targets", "context": "SFT TEACHER-PROBE NO-GO (deep teacher 6/120 faithful); fine-tuning line closed", "decision": "SFT of the 4B closed unless a teacher reaches ≥ 50 % faithfulness on distractor-rich targets", "alternatives": [ "train anyway" ], "rationale": "Haiku teacher 38 %, deep teacher 5 %", "consequence": "closed (R71)", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-481" ], "chapter": "walls", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-22" ] }, { "id": "D-094", "date": "2026-10-14", "time": null, "title": "Rule 29 experiment economics: Ceiling block (launch only if ≥ 2×MDE) and Budget block (smoke, wall caps, kill rule)", "context": "RULE 29 experiment economics: Ceiling + Budget + Transfer blocks; no Opus on dev sets", "decision": "Rule 29 experiment economics: Ceiling block (launch only if ≥ 2×MDE) and Budget block (smoke, wall caps, kill rule)", "alternatives": [ "free launches", "status quo", "Keep \"any Haiku Δ > 0 → Opus\"" ], "rationale": "≈ 1,440 avoidable Opus calls and long bad tests", "consequence": "reviewed, edits applied", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-885", "F-886", "F-889", "F-891" ], "chapter": "agent-lab", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-23", "D-method-25", "D-process-21", "D-results-14" ] }, { "id": "D-095", "date": "2026-10-14", "time": "19:56", "title": "LOOP v2 decisions D1–D12 (machine-side queue runners, smoke-first, STATE line, lanes, guards as scripts, handoff instead of compaction, daily brief, watchdog)", "context": "LOOP v2 retrospective: 133 ticks, ≈ 12.0k Claude calls (≈ 10.5k judge calls), 1 lever, ≈31 avoidable machine-hours", "decision": "LOOP v2 decisions D1–D12 (machine-side queue runners, smoke-first, STATE line, lanes, guards as scripts, handoff instead of compaction, daily brief, watchdog)", "alternatives": [ "keep v1 loop", "Fresh-context `claude -p` ticks (kept as option behind a shadow run)" ], "rationale": "26 incidents, ≈ 31 h avoidable machine time", "consequence": "T1–T7 built 10-15; T8 10-16", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-890", "F-927" ], "chapter": "agent-lab", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-24", "D-process-22" ] }, { "id": "D-097", "date": "2026-10-14", "time": "22:17", "title": "Publish research AND benchmark for other teams; publish v0.1 reference-answer text; v0.2/V-para refs stay hashes only", "context": "Open-source intent: research AND benchmark public; v0.1 reference-answer text made public", "decision": "Publish research AND benchmark for other teams; publish v0.1 reference-answer text; v0.2/V-para refs stay hashes only", "alternatives": [ "publish hashes only" ], "rationale": "other teams cannot run the judge comparably without refs", "consequence": "boar-bench carries refs with provenance on the dataset card", "reversible": "no (public)", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1310" ], "chapter": "reproduce", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-27" ] }, { "id": "D-098", "date": "2026-10-14", "time": null, "title": "Bounty chat gets method, packs and device proof only; numbers only after the shared rubric exists", "context": "RULE 29 experiment economics: Ceiling + Budget + Transfer blocks; no Opus on dev sets", "decision": "Bounty chat gets method, packs and device proof only; numbers only after the shared rubric exists", "alternatives": [ "post scores" ], "rationale": "numbers without a shared rubric are not comparable", "consequence": "post drafts carry no scores", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1058" ], "chapter": "packs", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-28" ] }, { "id": "D-099", "date": "2026-10-14", "time": null, "title": "Verdict, registry row (CLOSED.md) and exposure row in the same commit", "context": "Coordinator session dies on an expired login token; mini idle 2.2 h", "decision": "Verdict, registry row (CLOSED.md) and exposure row in the same commit", "alternatives": [ "separate bookkeeping", "Periodic backfill" ], "rationale": "ledger audit found 48 verdicts without registry rows", "consequence": "in force", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-887", "F-888" ], "chapter": "agent-lab", "visibility": "public", "status_label": "kept", "aliases": [ "D-method-26", "D-process-20" ] }, { "id": "D-100", "date": "2026-10-15", "time": "10:07", "title": "First public push by the coordinator, only for the two new repos, on the author's explicit instruction", "context": "Public repos live (first push by the coordinator, explicitly authorised for those two repos only)", "decision": "First public push by the coordinator, only for the two new repos, on the author's explicit instruction", "alternatives": [ "keep private" ], "rationale": "release", "consequence": "repos live; no-push rule stays for everything else", "reversible": "no", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-936" ], "chapter": "agent-lab", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-29" ] }, { "id": "D-101", "date": "2026-10-15", "time": "11:00", "title": "Accept the maintainer's native Kotlin app as the official submission; our numbers are labelled as the TypeScript engine's", "context": "Submission vehicle = maintainer's native Kotlin app; PR #85 (cut 7 engine as reference) open", "decision": "Accept the maintainer's native Kotlin app as the official submission; our numbers are labelled as the TypeScript engine's", "alternatives": [ "submit the RN app with PR #85", "the React Native app carrying cut 7 via PR" ], "rationale": "the maintainer owns the submission", "consequence": "claim v4.1 vehicle paragraph; native score pending", "reversible": "no (external)", "status": "superseded", "status_note": "On 2026-10-24 the author made BoarLite, a new closed app of his own, the bounty submission.", "superseded_by": "D-183", "verification_note": null, "fact_ids": [ "F-1238", "F-788" ], "chapter": "honest-number", "visibility": "public", "status_label": "superseded by D-183", "aliases": [ "D-chron-late-30", "D-system-16" ] }, { "id": "D-103", "date": "2026-10-15", "time": "21:34", "title": "Reject the lexicon numeric guard; ship calculator-only NUMGUARD r3", "context": "Independent reviewer REJECTs a numeric guard as unsafe; scope cut to calculator-only", "decision": "Reject the lexicon numeric guard; ship calculator-only NUMGUARD r3", "alternatives": [ "NUMGUARD-FINAL lexicon guard" ], "rationale": "confirmation review: a lexicon guard cannot be made safe (numeric health questions leak)", "consequence": "r3 pushed to PR #85", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1275" ], "chapter": "handoff", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-32" ] }, { "id": "D-104", "date": "2026-10-15", "time": "22:21", "title": "Lite tier is post-10-20 work, explored with quick diagnostic tests; starts with an unsafe-answer warning", "context": "Author: lite tier is an open, post-10-20 exploration line; starts with an unsafe-answer warning", "decision": "Lite tier is post-10-20 work, explored with quick diagnostic tests; starts with an unsafe-answer warning", "alternatives": [ "drop lite tier" ], "rationale": "2B −0.168 and 6/15 unsafe health answers; weaker phones matter", "consequence": "LITE2–LITE8, LITEREADERS etc. 10-15/16; closed for now", "reversible": "yes", "status": "superseded", "status_note": "On 2026-10-20 the small model became the main 8 GB engine, not a post-deadline lite line.", "superseded_by": "D-134", "verification_note": null, "fact_ids": [ "F-790" ], "chapter": "phone", "visibility": "public", "status_label": "superseded by D-134", "aliases": [ "D-chron-late-33" ] }, { "id": "D-105", "date": "2026-10-15", "time": null, "title": "Release as two fresh repos with one squashed commit, neutral author, curated whitelist + scrub + mechanical gate + independent review; never the private history", "context": "Public repos staged: scrub, mechanical gate, independent review (41/44 findings applied)", "decision": "Release as two fresh repos with one squashed commit, neutral author, curated whitelist + scrub + mechanical gate + independent review; never the private history", "alternatives": [ "Publish private repos" ], "rationale": "Private history holds held-out plaintext and personal data", "consequence": "Gate FAIL on 2 handle URLs fixed; 41/44 review findings applied; live 10:07", "reversible": "no (published)", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1312" ], "chapter": "reproduce", "visibility": "public", "status_label": "kept", "aliases": [ "D-process-24" ] }, { "id": "D-106", "date": "2026-10-15", "time": null, "title": "Drop the loop-operations memo and internal decision memo from the public repo; replace with a 9-line method summary", "context": "Author: lite tier is an open, post-10-20 exploration line; starts with an unsafe-answer warning", "decision": "Drop the loop-operations memo and internal decision memo from the public repo; replace with a 9-line method summary", "alternatives": [ "Publish as is" ], "rationale": "Reviewer: internal agent prompts, caps, people and channels", "consequence": "`method/METHOD.md`", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-930" ], "chapter": "agent-lab", "visibility": "public", "status_label": "kept", "aliases": [ "D-process-25" ] }, { "id": "D-107", "date": "2026-10-15", "time": null, "title": "Lite tier: thinking stays OFF; 1.7B Q4_K_M chosen over ternary 2B and Q8_0", "context": "Port spec of the health path written for the native app", "decision": "Lite tier: thinking stays OFF; 1.7B Q4_K_M chosen over ternary 2B and Q8_0", "alternatives": [ "Bonsai ternary 2B/4B; Q8_0" ], "rationale": "LITE2 2B −0.057; LITE-THINK flat/negative; QUANTLITE flat and slower", "consequence": "Lite tier remains a product idea, not a claim", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-645" ], "chapter": "deep-tier", "visibility": "public", "status_label": "kept", "aliases": [ "D-results-15" ] }, { "id": "D-108", "date": "2026-10-15", "time": null, "title": "Keep the native-app comparison internal until the final build is scored", "context": "Closing-sprint loop on in the live session", "decision": "Keep the native-app comparison internal until the final build is scored", "alternatives": [ "publish now", "publish" ], "rationale": "intermediate build, decision by the author", "consequence": "Number withheld from claim text", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1237", "F-1263" ], "chapter": "honest-number", "visibility": "public", "status_label": "kept", "aliases": [ "D-results-18", "D-system-21" ] }, { "id": "D-109", "date": "2026-10-15", "time": null, "title": "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 c", "context": "NUMGUARD r1→r3: lexicon guard rejected as unsafe by review; calculator-only r3 PASS", "decision": "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)", "alternatives": [ "model with a disclaimer" ], "rationale": "safety; model answered doses from memory on probes", "consequence": null, "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": "citation fix: 'never reach the 4B' is l.21. AI-reviewed, not clinically reviewed", "fact_ids": [ "F-374" ], "chapter": "builds", "visibility": "public", "status_label": "kept", "aliases": [ "D-system-18" ] }, { "id": "D-110", "date": "2026-10-16", "time": "03:39", "title": "TK2-FIX is a correctness fix, not a score lever; out of PR #85 before 10-20", "context": "Critique raises the one-seed noise rule from 0.03 to ≥ 0.05", "decision": "TK2-FIX is a correctness fix, not a score lever; out of PR #85 before 10-20", "alternatives": [ "include in PR" ], "rationale": "+0.010 pooled, needs a builder change too", "consequence": "queued as a post-10-20 patch", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1284" ], "chapter": "handoff", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-34" ] }, { "id": "D-111", "date": "2026-10-16", "time": "09:4x", "title": "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 ot", "context": "Judge authentication expired; no judge passes for ≈ 2.5 h; lanes idle ≈ 5 h (coordinator fault acknowledged)", "decision": "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\"", "alternatives": [ "idle waiting" ], "rationale": "≈ 5 h of idle lanes on 10-16", "consequence": "settled the critique's point on 3-minute filler probes", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1199" ], "chapter": "measurer", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-35" ] }, { "id": "D-112", "date": "2026-10-16", "time": null, "title": "deps.warn (unsafe-answer warning) stays off by default", "context": null, "decision": "deps.warn (unsafe-answer warning) stays off by default", "alternatives": [ "default-on" ], "rationale": "out-of-sample recall 3/8 (lexicon) and 12/18 with FP 8/92 (detector)", "consequence": "not handed to the PR", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-389" ], "chapter": "builds", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-36" ] }, { "id": "D-113", "date": "2026-10-16", "time": "18:28", "title": "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", "context": "Adversarial critique: lite \"= 4B\" overstated (kit −0.111), lite gap is reading too (−0.070 at identical contexts); closed loop risk", "decision": "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\"", "alternatives": [ "keep wording" ], "rationale": "over-stated internal summaries", "consequence": "7 re-wordings applied", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1201" ], "chapter": "measurer", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-37" ] }, { "id": "D-115", "date": "2026-10-16", "time": null, "title": "Claude judge stays the headline (lower of two judges); judge family named next to every ratio; win/tie/loss leads", "context": "Claim v4.2 / v4.2.1: judge family named on every ratio; native comparison kept internal", "decision": "Claude judge stays the headline (lower of two judges); judge family named next to every ratio; win/tie/loss leads", "alternatives": [ "headline the higher cross-family number", "Report the higher judge", "lead with OpenAI-judge numbers (deep crosses 0.50)" ], "rationale": "analysis-plan rule: when judges differ by > 0.05, the lower is the headline", "consequence": "claim v4.2.x", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1189", "F-1239", "F-1240" ], "chapter": "honest-number", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-39", "D-process-27", "D-results-17" ] }, { "id": "D-116", "date": "2026-10-16", "time": null, "title": "Apply 7 re-wordings from the adversarial critique (retract \"gap is reach, not reading\", etc.)", "context": "Independent adversarial CRITIQUE: 7 re-wordings applied", "decision": "Apply 7 re-wordings from the adversarial critique (retract \"gap is reach, not reading\", etc.)", "alternatives": [ "Defend summaries" ], "rationale": "Numbers real, summaries scope-stripped", "consequence": "Corrections logged as QUEUE rows", "reversible": "no", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1188" ], "chapter": "measurer", "visibility": "public", "status_label": "kept", "aliases": [ "D-process-26" ] }, { "id": "D-117", "date": "2026-10-16", "time": null, "title": "NUMGUARD lexicon guard rejected as unsafe; calculator-only r3 kept", "context": "NUMGUARD r3 + HEALTHGAP r2 added to the upstream PR; WARN banner spec v2", "decision": "NUMGUARD lexicon guard rejected as unsafe; calculator-only r3 kept", "alternatives": [ "ship the lexicon guard (passed tuned gates)" ], "rationale": "a lexicon that lets the model answer a detector-fired health turn cannot be made safe", "consequence": "r3 shipped to PR #85", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1278" ], "chapter": "handoff", "visibility": "public", "status_label": "kept", "aliases": [ "D-results-16" ] }, { "id": "D-118", "date": "2026-10-16", "time": null, "title": "Lite tier proposed for 4–6 GB phones: Qwen3-1.7B behind the full safety engine", "context": "Lite tier (Qwen3-1.7B) evaluated; LITESAFETY PASS; critique corrects \"1.7B = 4B\"", "decision": "Lite tier proposed for 4–6 GB phones: Qwen3-1.7B behind the full safety engine", "alternatives": [ "Gemma-3-1B, LFM2-1.2B (worse readers), Q8 quant (flat)" ], "rationale": "1.1 GB file, RSS ≈ 2.3 GB; safety routing identical to 4B", "consequence": "offered as option; quality wording corrected after critique", "reversible": "yes", "status": "superseded", "status_note": "On 2026-10-20 Qwen3-1.7B became the main 8 GB engine; the 4B became the optional tier.", "superseded_by": "D-134", "verification_note": null, "fact_ids": [ "F-1190" ], "chapter": "measurer", "visibility": "public", "status_label": "superseded by D-134", "aliases": [ "D-system-19" ] }, { "id": "D-119", "date": "2026-10-16", "time": null, "title": "Expert streaming (LES fork) supersedes earlier loader/kernel candidates on paper for a deep-tier device plan", "context": "LESPORT: expert-streaming fork runs the 35B in 2.8 GiB RSS on desktop CPU, judged parity TRANSFER", "decision": "Expert streaming (LES fork) supersedes earlier loader/kernel candidates on paper for a deep-tier device plan", "alternatives": [ "bmoe (7.9 GB on phone)" ], "rationale": "2.8 GiB RSS, judged parity", "consequence": "phone flash speed unmeasured; Android buffer unbuilt", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-792" ], "chapter": "phone", "visibility": "public", "status_label": "kept", "aliases": [ "D-system-20" ] }, { "id": "D-120", "date": "2026-10-17", "time": "20:59", "title": "Claim v4.2.4: state that the submitted app ports retrieval and prompts, not the health path or deterministic answers", "context": "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", "decision": "Claim v4.2.4: state that the submitted app ports retrieval and prompts, not the health path or deterministic answers", "alternatives": [ "keep the earlier sentence", "earlier wording \"port of the rules\"" ], "rationale": "maintainer's own portal report + MAINT-V01", "consequence": "claim wording as of 2026-10-17; the claim text stays open until closure (open front OF-5)", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1244", "F-1292" ], "chapter": "handoff", "visibility": "public", "status_label": "kept", "aliases": [ "D-chron-late-40", "D-system-22" ] }, { "id": "D-122", "date": "2026-10-17", "time": null, "title": "When judges from two families disagree, the lower number is the headline; judge family named next to every ratio", "context": "Claim v4.2.3: \"the lower judge is the headline\"; judge family named next to every ratio", "decision": "When judges from two families disagree, the lower number is the headline; judge family named next to every ratio", "alternatives": [ "lead with the cross-family number" ], "rationale": "XFAM offset ≈ +0.10", "consequence": "claim v4.2.3", "reversible": "yes", "status": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1243" ], "chapter": "honest-number", "visibility": "public", "status_label": "kept", "aliases": [ "D-method-27" ] }, { "id": "D-123", "date": "2026-10-18", "time": "16:07", "title": "D1: no LFM2.5 desktop replay", "context": "decided by: delegated coordinator (author delegated)", "decision": "D1: no LFM2.5 desktop replay", "alternatives": [ "run replay ~1.5 h" ], "rationale": "audit already 7/58 vs 3/73 wrong; product already chose Qwen", "consequence": "closed", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": "Verification note: The parenthetical wording ('scored internally, final build published when it ships') is the claim text's, sourced at not . 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).", "fact_ids": [ "F-1370", "F-1371", "F-1372", "F-1373", "F-1374", "F-1375", "F-1833" ], "chapter": "small-engine", "visibility": "public", "aliases": [ "D-u-1820-01", "D-u-1820-02", "D-u-1820-03", "D-u-1820-04", "D-u-1820-05", "D-u-1820-06", "D-u-results-02" ] }, { "id": "D-124", "date": "2026-10-18", "time": "16:07", "title": "D7: upload the research site only as v1.0 at closure; v0.x are internal build states", "context": "decided by: delegated coordinator", "decision": "D7: upload the research site only as v1.0 at closure; v0.x are internal build states", "alternatives": [ "upload v0.1 now" ], "rationale": "one publication, no intermediate states", "consequence": "folded into publication PLAN", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1376" ], "chapter": "closure", "visibility": "public", "aliases": [ "D-u-1820-07" ] }, { "id": "D-125", "date": "2026-10-18", "time": "05:02", "title": "Recommend the honest Detailed prompt D1, then D4 (D1 + capability rule) as the Detailed prompt; no Succinct-honest", "context": "decided by: coordinator under frozen rule", "decision": "Recommend the honest Detailed prompt D1, then D4 (D1 + capability rule) as the Detailed prompt; no Succinct-honest", "alternatives": [ "D2, S1, S3" ], "rationale": "wrong claims down, ratio within -0.03 rule, refusals 1-2/49; S1 refuses 7/49", "consequence": "sent to maintainer via liaison", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1377", "F-1832" ], "chapter": "submission", "visibility": "public", "aliases": [ "D-u-1820-08", "D-u-results-01" ] }, { "id": "D-126", "date": "2026-10-18", "time": null, "title": "Never ship a capability sentence without health forced to guard texts", "context": "decided by: coordinator", "decision": "Never ship a capability sentence without health forced to guard texts", "alternatives": [ "ship L1" ], "rationale": "L1 line added wrong first-aid steps on first-aid chats", "consequence": "rule standing", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1378" ], "chapter": "small-engine", "visibility": "public", "aliases": [ "D-u-1820-09" ] }, { "id": "D-127", "date": "2026-10-19", "time": "13:30", "title": "Focus finishing work on \"tests with sources\" (user has the matching pack)", "context": "decided by: author", "decision": "Focus finishing work on \"tests with sources\" (user has the matching pack)", "alternatives": [ "continue no-source prompt tests (PROMPT3M)" ], "rationale": "bounty finish", "consequence": "PACKREAD run; PROMPT3M paused, later closed", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1379" ], "chapter": "delivery", "visibility": "public", "aliases": [ "D-u-1820-10" ] }, { "id": "D-128", "date": "2026-10-19", "time": "14:27", "title": "Release saferoute patch despite SAFEPACK frozen-rule FAIL (app ships with a warning); FAIL kept on record", "context": "decided by: author", "decision": "Release saferoute patch despite SAFEPACK frozen-rule FAIL (app ships with a warning); FAIL kept on record", "alternatives": [ "hold patch" ], "rationale": "rule judged too strict for stage", "consequence": "reversed same day: held out of the pushed stack after the direction review", "reversible": "yes", "status": "reversed", "status_label": "reversed", "status_note": "Reversed the same day: the saferoute patch was held out of the pushed stack after the independent direction review.", "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1380", "F-1834" ], "chapter": "delivery", "visibility": "public", "aliases": [ "D-u-1820-11", "D-u-results-03" ] }, { "id": "D-129", "date": "2026-10-19", "time": "15:19", "title": "Stop opening new arms; consolidate; hold saferoute; correct over-reads (SMALLCLS, SAFEPACK)", "context": "decided by: independent reviewer, adopted by coordinator", "decision": "Stop opening new arms; consolidate; hold saferoute; correct over-reads (SMALLCLS, SAFEPACK)", "alternatives": [ "open SMALLCLS-2", "agentic search loop" ], "rationale": "scope drift two days before closure; small-n in-sample results drove release calls", "consequence": "adopted for saferoute; new arms nonetheless resumed 10-20 under new program", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1381" ], "chapter": "delivery", "visibility": "public", "aliases": [ "D-u-1820-12" ] }, { "id": "D-130", "date": "2026-10-19", "time": "14:36", "title": "Resume three safety-held items: per-answer health caution ON, drinking-water card, chat routing L2+L3 without L1", "context": "decided by: author", "decision": "Resume three safety-held items: per-answer health caution ON, drinking-water card, chat routing L2+L3 without L1", "alternatives": [ "keep held" ], "rationale": "app ships with a warning; harm came from L1 line only", "consequence": "in the pushed PR stack", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1382" ], "chapter": "small-engine", "visibility": "public", "aliases": [ "D-u-1820-13" ] }, { "id": "D-131", "date": "2026-10-19", "time": "23:08", "title": "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", "context": "decided by: coordinator from 8-agent workflow (both critics converged)", "decision": "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", "alternatives": [ "keep old bench as headline" ], "rationale": "five misreadings of the bounty in the old bench", "consequence": "led to v0.4 items + spec v1", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1383", "F-1836" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-1820-14", "D-u-results-05" ] }, { "id": "D-132", "date": "2026-10-20", "time": null, "title": "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", "context": "decided by: coordinator, author direction", "decision": "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", "alternatives": [ "r1 quotas (72 items, >= 50 % travel)" ], "rationale": "match the inspiring post's queries", "consequence": "applied to v0.4", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1384" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-1820-15" ] }, { "id": "D-133", "date": "2026-10-20", "time": "16:45", "title": "Resolve 5 disputed points; drop one item contradicted by both families; tag 4 V-exact anchors out of the headline; freeze v0.4", "context": "decided by: author", "decision": "Resolve 5 disputed points; drop one item contradicted by both families; tag 4 V-exact anchors out of the headline; freeze v0.4", "alternatives": [ "keep disputed items" ], "rationale": "both-family fact check", "consequence": "frozen, sha f9ad2f2f...", "reversible": "no (frozen)", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1385", "F-1838" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-1820-16", "D-u-results-07" ] }, { "id": "D-134", "date": "2026-10-20", "time": "11:50", "title": "Qwen3-1.7B = main 8 GB engine; 4B = optional \"more complete\" tier", "context": "decided by: author", "decision": "Qwen3-1.7B = main 8 GB engine; 4B = optional \"more complete\" tier", "alternatives": [ "keep 4B default", "Qwen2.5-1.5B (faster but weaker)" ], "rationale": "Pixel 8a: 1.7B 13-16 tok/s, answers 13-16 s; 4B 3.6-6 tok/s, 29-46 s, one LMK kill", "consequence": "adopted; confirmed at 19:07 grill", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1386", "F-1837" ], "chapter": "small-engine", "visibility": "public", "aliases": [ "D-u-1820-17", "D-u-results-06" ] }, { "id": "D-135", "date": "2026-10-20", "time": "17:35", "title": "Propose to maintainer: 8 GB = Qwen3-1.7B + fixed warning line instead of compact declines", "context": "decided by: author (send OK)", "decision": "Propose to maintainer: 8 GB = Qwen3-1.7B + fixed warning line instead of compact declines", "alternatives": [ "keep declines" ], "rationale": "CMP30OFF: 1.7B with declines released leads 1.5B", "consequence": "superseded at 20:24 by refuse-by-default", "reversible": "yes", "status": "superseded", "status_label": "superseded by D-149", "status_note": "superseded at 20:24 by refuse-by-default", "superseded_by": "D-149", "verification_note": null, "fact_ids": [ "F-1387" ], "chapter": "submission", "visibility": "public", "aliases": [ "D-u-1820-18" ] }, { "id": "D-136", "date": "2026-10-20", "time": "19:07", "title": "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 ", "context": "decided by: author", "decision": "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", "alternatives": [ "phone A/B of 4B", "C before bounty" ], "rationale": "LITEPACKS NO CHANGE", "consequence": "adopted", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1388" ], "chapter": "delivery", "visibility": "public", "aliases": [ "D-u-1820-19" ] }, { "id": "D-137", "date": "2026-10-20", "time": "19:07", "title": "Ruler headline weights the 11 health_safety items at 0.5, adds a safety gate, publishes weight-1 U beside it", "context": "decided by: author", "decision": "Ruler headline weights the 11 health_safety items at 0.5, adds a safety gate, publishes weight-1 U beside it", "alternatives": [ "weight 1" ], "rationale": "bounty asks for casual lookup/research, never names health", "consequence": "in spec v1 §2; reviewer F3 flags timing (decided after items existed)", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1389" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-1820-20" ] }, { "id": "D-138", "date": "2026-10-20", "time": "20:24", "title": "Ruler headline = coverage x U decomposition + U-total; product refuses by default when no source is kept, with a \"show anyway\" control", "context": "decided by: author", "decision": "Ruler headline = coverage x U decomposition + U-total; product refuses by default when no source is kept, with a \"show anyway\" control", "alternatives": [ "answer under a warning line" ], "rationale": "separate what packs answer from memory answers", "consequence": "spec v1 r2", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1390", "F-1840" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-1820-21", "D-u-results-09" ] }, { "id": "D-139", "date": "2026-10-20", "time": "20:39", "title": "Verdict arm = A∪ (all persona packs installed once, same for every item); per-family arm A kept as a labelled row", "context": "decided by: author on reviewer finding F1", "decision": "Verdict arm = A∪ (all persona packs installed once, same for every item); per-family arm A kept as a labelled row", "alternatives": [ "arm A or arm B as verdict" ], "rationale": "per-family arm = oracle topic routing", "consequence": "spec v1 r2", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": "Verification note: Reviewer finding F1 is at (TL;DR 1) and :14 (F1 heading); line 7 is the decline-arm finding.", "fact_ids": [ "F-1391" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-1820-22" ] }, { "id": "D-140", "date": "2026-10-20", "time": "21:45", "title": "Persona map v1: travel -> wikivoyage + visa (+ area tile); tech -> crypto; science -> medlineplus + firstaid; place -> factbook; humanities -> none; NIST pack e", "context": "decided by: coordinator draft + independent approver (6 edits)", "decision": "Persona map v1: travel -> wikivoyage + visa (+ area tile); tech -> crypto; science -> medlineplus + firstaid; place -> factbook; humanities -> none; NIST pack excluded", "alternatives": [ "include NIST, simplewiki" ], "rationale": "chosen from pack-card domain lines only", "consequence": "not frozen, waits author veto", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1392", "F-1841" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-1820-23", "D-u-results-10" ] }, { "id": "D-141", "date": "2026-10-20", "time": "22:29", "title": "GATE v2: screen bar = max measured null net + 2", "context": "decided by: coordinator", "decision": "GATE v2: screen bar = max measured null net + 2", "alternatives": [ "fixed bar 6" ], "rationale": "null runs fixed 23-31 % of problems; old 6-fix bar sat at the null", "consequence": "used for overnight screens", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1393" ], "chapter": "small-engine", "visibility": "public", "aliases": [ "D-u-1820-24" ] }, { "id": "D-142", "date": "2026-10-21", "time": null, "title": "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", "context": "decided by: delegated coordinator (proposal, reviewed)", "decision": "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", "alternatives": [ "Gemma default", "4B default" ], "rationale": "Gemma +0.029 at 1.5× time +1.4 GB; Ministral +0.062 at ~3× time; 4B +0.076 at ~2.5× time", "consequence": "stood; superseded in practice by RC1 shipping 1.5B", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1527" ], "chapter": "submission", "visibility": "public", "aliases": [ "D-u-2122-01" ] }, { "id": "D-143", "date": "2026-10-21", "time": "02:38", "title": "Gemma 4 E2B only as opt-in catalog entry with thinking guards (thinking off, budget 0, n_predict cap, post-filter); exact file pinned", "context": "decided by: delegated coordinator", "decision": "Gemma 4 E2B only as opt-in catalog entry with thinking guards (thinking off, budget 0, n_predict cap, post-filter); exact file pinned", "alternatives": [ "Gemma default", "drop Gemma" ], "rationale": "phone quality = 1.7B, 1.5-2.8× time, +1.1-1.4 GB; runtime risks (thinking default on, KV growth)", "consequence": "proposal; Gemma not in RC1 catalog so moot on RC1", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1528" ], "chapter": "submission", "visibility": "public", "aliases": [ "D-u-2122-02" ] }, { "id": "D-144", "date": "2026-10-21", "time": "12:45", "title": "RULING R1: judge = judge.v1 unchanged + DECLINE2STAGE zero per detected non-answer; D 0.058 kept as sensitivity row", "context": "decided by: delegated coordinator (override of a frozen FAIL, accepted with conditions by reviewer)", "decision": "RULING R1: judge = judge.v1 unchanged + DECLINE2STAGE zero per detected non-answer; D 0.058 kept as sensitivity row", "alternatives": [ "judge v1.1", "decline subtraction only" ], "rationale": "judge v1.1/v1.1b changed CE verdicts (5-7/40 flips); DECLINE2STAGE's 2 bar-(b) hits were engine refusal strings", "consequence": "adopted into ruler r3/r4", "reversible": "yes (pre-exposure)", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1529", "F-1845" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-2122-03", "D-u-results-14" ] }, { "id": "D-145", "date": "2026-10-21", "time": "13:12", "title": "GOAL: two numbers U_8GB (+0.03 Opus, CE-shown must not rise) and T_8GB (7.6 → 5.0 s)", "context": "decided by: author", "decision": "GOAL: two numbers U_8GB (+0.03 Opus, CE-shown must not rise) and T_8GB (7.6 → 5.0 s)", "alternatives": [ "open exploration" ], "rationale": "focus the last days on measurable engine gains", "consequence": "T_8GB re-based to the native app same day; U_8GB lines closed at nulls", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1530" ], "chapter": "submission", "visibility": "public", "aliases": [ "D-u-2122-04" ] }, { "id": "D-146", "date": "2026-10-21", "time": "13:13", "title": "Winning thesis = honest 8 GB product + open ruler", "context": "decided by: author (GRILL)", "decision": "Winning thesis = honest 8 GB product + open ruler", "alternatives": [ "compete on the deep tier" ], "rationale": "35B not deliverable on 8 GB in time", "consequence": "stood until 10-22 pivot reframed it", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1531" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-2122-05" ] }, { "id": "D-147", "date": "2026-10-21", "time": "13:13", "title": "35B deep mode not delivered; deep tier appears only as roadmap with desktop numbers", "context": "decided by: author (GRILL)", "decision": "35B deep mode not delivered; deep tier appears only as roadmap with desktop numbers", "alternatives": [ "12 GB deep tier" ], "rationale": "focus on 8 GB UX; no Android expert-streaming build", "consequence": "stood", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1532" ], "chapter": "deep-tier", "visibility": "public", "aliases": [ "D-u-2122-06" ] }, { "id": "D-148", "date": "2026-10-21", "time": "13:13", "title": "8 GB default model: Qwen3-1.7B, 4B optional (sent to maintainer only within the short package)", "context": "decided by: 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).", "decision": "8 GB default model: Qwen3-1.7B, 4B optional (sent to maintainer only within the short package)", "alternatives": null, "rationale": null, "consequence": null, "reversible": null, "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": "Verification note: Faithful to GRILL D3 but compares a 73-question all-items p50 for the 1.7B (4.2 s) with 9-question model-path answer ranges for the 4B (29-46 s). Same 73 set, same phone: 1.7B p50 4.2 s vs 4B p50 25.5 s (cut 7, 10-20).", "fact_ids": [ "F-1533" ], "chapter": "submission", "visibility": "public", "aliases": [ "D-u-2122-07" ] }, { "id": "D-149", "date": "2026-10-21", "time": "13:13", "title": "No source found → refuse by default + \"show anyway\", shown answers carry a warning line", "context": "decided by: author (GRILL)", "decision": "No source found → refuse by default + \"show anyway\", shown answers carry a warning line", "alternatives": [ "always answer" ], "rationale": "honesty on 8 GB", "consequence": "later narrowed: show-anyway only behind second confirmation, never on health (REVEAL-CENSUS); RC1 has no warning line", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1534", "F-1551" ], "chapter": "submission", "visibility": "public", "aliases": [ "D-u-2122-08", "D-u-2122-25" ] }, { "id": "D-150", "date": "2026-10-21", "time": "13:13", "title": "Article = research site on a project subdomain built from the public research repo; story = timeline of decisions, not a score line", "context": "decided by: author (GRILL)", "decision": "Article = research site on a project subdomain built from the public research repo; story = timeline of decisions, not a score line", "alternatives": [ "pages on existing site" ], "rationale": "full control of look; avoid cross-set score charts", "consequence": "stands (this publication)", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1535" ], "chapter": "closure", "visibility": "public", "aliases": [ "D-u-2122-09" ] }, { "id": "D-151", "date": "2026-10-21", "time": "13:13", "title": "Final score = submitted native APK on Pixel 8a, v0.4 frozen (75 items), Claude + OpenAI gpt-5.6-sol judges, frontier + web references", "context": "decided by: author (GRILL)", "decision": "Final score = submitted native APK on Pixel 8a, v0.4 frozen (75 items), Claude + OpenAI gpt-5.6-sol judges, frontier + web references", "alternatives": [ "engine-only score" ], "rationale": "the bounty judges the app", "consequence": "ruler r3/r4 built for it; run pending a new RC", "reversible": "partly (depends on RC)", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1536" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-2122-10" ] }, { "id": "D-152", "date": "2026-10-21", "time": "15:5x", "title": "Submission = the maintainer's native Kotlin app, not the RN bench engine", "context": "decided by: author", "decision": "Submission = the maintainer's native Kotlin app, not the RN bench engine", "alternatives": [ "engine-only claim" ], "rationale": "that is what will be submitted", "consequence": "re-based speed lines (bench levers page-only) and claim section 0", "reversible": "no (fact of the submission)", "status": "superseded", "status_label": "superseded by D-183", "status_note": "re-based speed lines (bench levers page-only) and claim section 0", "superseded_by": "D-183", "verification_note": null, "fact_ids": [ "F-1537" ], "chapter": "submission", "visibility": "public", "aliases": [ "D-u-2122-11" ] }, { "id": "D-153", "date": "2026-10-21", "time": "16:01", "title": "No 512-token cap ask (CAP512 not a lever)", "context": "decided by: delegated coordinator (frozen rule)", "decision": "No 512-token cap ask (CAP512 not a lever)", "alternatives": [ "cap ask in NOTE-1" ], "rationale": "saves 0 s p50, cuts 10.9 % mid-sentence", "consequence": "dropped", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1538" ], "chapter": "small-engine", "visibility": "public", "aliases": [ "D-u-2122-12" ] }, { "id": "D-156", "date": "2026-10-21", "time": "20:28", "title": "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", "context": "decided by: author", "decision": "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", "alternatives": [ "verdict on app without packs" ], "rationale": "thesis \"app + packs answers well\"; two families", "consequence": "in r3/r4", "reversible": "yes (pre-run)", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1541", "F-1852" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-2122-15", "D-u-results-21" ] }, { "id": "D-157", "date": "2026-10-21", "time": "20:48", "title": "Freeze RULER-SPEC v1 r3 (private)", "context": "decided by: delegated coordinator + reviewer ACCEPT WITH EDITS", "decision": "Freeze RULER-SPEC v1 r3 (private)", "alternatives": [ "keep binary checklist judge (NO-GO earlier)" ], "rationale": "pre-registration before any v0.4 answer", "consequence": "frozen sha 0fac2286", "reversible": "no (frozen)", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1542" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-2122-16" ] }, { "id": "D-158", "date": "2026-10-21", "time": "21:18", "title": "Freeze r4: add information-available split (FACT-AVAIL) + ORACLE/ORACLE-0 rows; U-total stays verdict", "context": "decided by: author asked", "decision": "Freeze r4: add information-available split (FACT-AVAIL) + ORACLE/ORACLE-0 rows; U-total stays verdict", "alternatives": [ "no split" ], "rationale": "separate \"packs lack it\" from \"app fails to deliver it\"", "consequence": "frozen sha 82772e33; labels quarantined", "reversible": "no (frozen)", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1543", "F-1853" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-2122-17", "D-u-results-22" ] }, { "id": "D-159", "date": "2026-10-21", "time": "22:10", "title": "Close decoder-level grounding axis on the 1.7B", "context": "decided by: delegated coordinator (frozen gate)", "decision": "Close decoder-level grounding axis on the 1.7B", "alternatives": [ "AdaCAD (unopened)" ], "rationale": "all arms net 4-6 vs bar 9", "consequence": "closed", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1544" ], "chapter": "small-engine", "visibility": "public", "aliases": [ "D-u-2122-18" ] }, { "id": "D-160", "date": "2026-10-22", "time": "01:10", "title": "Carry chunk vectors forward only in no-displacement \"extra\" mode; fusion harmful on real items", "context": "decided by: delegated coordinator", "decision": "Carry chunk vectors forward only in no-displacement \"extra\" mode; fusion harmful on real items", "alternatives": [ "fusion" ], "rationale": "fusion lost 6-8/22 real items", "consequence": "G2 full kit then failed real-item recovery (3/25)", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1545" ], "chapter": "delivery", "visibility": "public", "aliases": [ "D-u-2122-19" ] }, { "id": "D-161", "date": "2026-10-22", "time": "16:05", "title": "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", "context": "decided by: author", "decision": "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", "alternatives": [ "keep model/engine levers" ], "rationale": "QFORM 0.88 vs 0.19", "consequence": "two lanes (retrieval, coverage); NOTE-5 first ask", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1546", "F-1854" ], "chapter": "delivery", "visibility": "public", "aliases": [ "D-u-2122-20", "D-u-results-23" ] }, { "id": "D-162", "date": "2026-10-22", "time": "17:3x", "title": "Small focused packs, not giant ones; full Wikipedia = negative control", "context": "decided by: author", "decision": "Small focused packs, not giant ones; full Wikipedia = negative control", "alternatives": [ "full Wikipedia" ], "rationale": "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", "consequence": "composition (merge) becomes the central question", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1547" ], "chapter": "delivery", "visibility": "public", "aliases": [ "D-u-2122-21" ] }, { "id": "D-163", "date": "2026-10-22", "time": null, "title": "Composer UX = one options button + chips only for non-default settings; auto search-term chips dropped", "context": "decided by: author", "decision": "Composer UX = one options button + chips only for non-default settings; auto search-term chips dropped", "alternatives": [ "4-chip query bar MVP" ], "rationale": "T1 term extraction failed (25 → 23 %)", "consequence": "design adopted for the fork", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1548" ], "chapter": "delivery", "visibility": "public", "aliases": [ "D-u-2122-22" ] }, { "id": "D-164", "date": "2026-10-22", "time": "22:41", "title": "Withdraw F3 (health no-source fallback) from the search fixes", "context": "decided by: delegated coordinator (frozen safety rule)", "decision": "Withdraw F3 (health no-source fallback) from the search fixes", "alternatives": [ "keep F3 for +0.010" ], "rationale": "unsafe 9 → 13 on 65 probes", "consequence": "shippable set = gidf + F1 + F2 + F4 (+0.013, not a lever)", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1549", "F-1855" ], "chapter": "delivery", "visibility": "public", "aliases": [ "D-u-2122-23", "D-u-results-24" ] }, { "id": "D-165", "date": "2026-10-22", "time": "22:09", "title": "Delivery is measured at chunk level from now on", "context": "decided by: delegated coordinator", "decision": "Delivery is measured at chunk level from now on", "alternatives": [ "article-level" ], "rationale": "article-level overstated F1b by 13 facts", "consequence": "standing method", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1550", "F-1856" ], "chapter": "delivery", "visibility": "public", "aliases": [ "D-u-2122-24", "D-u-results-25" ] }, { "id": "D-167", "date": "2026-10-23", "time": "11:53", "title": "Close the research lines; no new arms after PLUGBATCH-SELF and STYLE-SEFIX; the rest of the work is acceptance tests", "context": "decided by: author", "decision": "Close the research lines; no new arms after PLUGBATCH-SELF and STYLE-SEFIX; the rest of the work is acceptance tests", "alternatives": [ "keep running levers" ], "rationale": "\"we reached where we needed to\"; ensure the app delivers it", "consequence": "Held for search/decoder/detector lines; reopened on the content side at 15:39 (D-u-2324-08)", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1687", "F-1857" ], "chapter": "closure", "visibility": "public", "aliases": [ "D-u-2324-01", "D-u-results-26" ] }, { "id": "D-168", "date": "2026-10-23", "time": "11:53", "title": "Adopted for delivery: SEFIX-124 search stack + W78 reading-cost stack; Succinct for models ≤ 2 GB; Deep Research hidden; detector v3b stays; decoder/readers/gat", "context": "decided by: author (plan by coordinator)", "decision": "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", "alternatives": [ "Detailed default, Deep mode, detector retrains" ], "rationale": "measured results of 10-22/23", "consequence": "Carried into ENGINE-RC", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1688" ], "chapter": "closure", "visibility": "public", "aliases": [ "D-u-2324-02" ] }, { "id": "D-169", "date": "2026-10-23", "time": "11:58", "title": "Closing tests run on the desktop twin with no phone", "context": "decided by: author", "decision": "Closing tests run on the desktop twin with no phone", "alternatives": [ "phone acceptance days" ], "rationale": "phone not available", "consequence": "UXMATRIX, U5, PACKPROOF ran on desktop", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1689" ], "chapter": "closure", "visibility": "public", "aliases": [ "D-u-2324-03" ] }, { "id": "D-170", "date": "2026-10-23", "time": "12:19", "title": "Thesis to present with the app: right whenever the content is in a pack; app built for it", "context": "decided by: author", "decision": "Thesis to present with the app: right whenever the content is in a pack; app built for it", "alternatives": [ "absolute accuracy claims" ], "rationale": "product framing", "consequence": "PACKPROOF supports only a conditional form (65 % when delivered)", "reversible": "yes (framing)", "status": "superseded", "status_label": "superseded by D-215", "status_note": "restated by the author on 2026-10-25 (D-215); the closing tests supported this sentence only in a conditional form", "superseded_by": "D-215", "verification_note": null, "fact_ids": [ "F-1690" ], "chapter": "closure", "visibility": "public", "aliases": [ "D-u-2324-04" ] }, { "id": "D-171", "date": "2026-10-23", "time": "13:20", "title": "Defaults by rule: Succinct stays, Deep hidden, no change to the honesty surface", "context": "decided by: delegated coordinator (frozen rule)", "decision": "Defaults by rule: Succinct stays, Deep hidden, no change to the honesty surface", "alternatives": [ "Detailed default" ], "rationale": "UXMATRIX matrix", "consequence": "Confirmed later by D-u-2324-13", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1691" ], "chapter": "closure", "visibility": "public", "aliases": [ "D-u-2324-05" ] }, { "id": "D-172", "date": "2026-10-23", "time": "14:58", "title": "Keep Wikipedia packs in the default kit K13", "context": "decided by: delegated coordinator (frozen gate), asked by author", "decision": "Keep Wikipedia packs in the default kit K13", "alternatives": [ "NOWIKI", "NOWIKI2 kits" ], "rationale": "removing them loses 9 MAIN facts, 10 general items, 2 health sources", "consequence": "K13 stays", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1692", "F-1859" ], "chapter": "delivery", "visibility": "public", "aliases": [ "D-u-2324-06", "D-u-results-28" ] }, { "id": "D-173", "date": "2026-10-23", "time": "14:39", "title": "No new packs for coverage (gaps are search, not missing content)", "context": "decided by: delegated coordinator, asked by author", "decision": "No new packs for coverage (gaps are search, not missing content)", "alternatives": [ "build philosophy", "blog", "Gutenberg packs" ], "rationale": "44/50 gaps already in K13; candidate packs ≈ 1 fact each", "consequence": "No pack builds", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1693" ], "chapter": "delivery", "visibility": "public", "aliases": [ "D-u-2324-07" ] }, { "id": "D-174", "date": "2026-10-23", "time": "15:39", "title": "CONTENT-FIRST: no model change; make the right chunk reach Qwen3-1.7B; build/shape the packs", "context": "decided by: author", "decision": "CONTENT-FIRST: no model change; make the right chunk reach Qwen3-1.7B; build/shape the packs", "alternatives": [ "model change, bigger model" ], "rationale": "\"even the small model should answer right having the content\"", "consequence": "PACKSHAPE lever-candidate; others NO", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1694" ], "chapter": "closure", "visibility": "public", "aliases": [ "D-u-2324-08" ] }, { "id": "D-175", "date": "2026-10-23", "time": "17:07", "title": "Shaped factbook v2: card drafted as user-selected only; publish waits for author OK", "context": "decided by: delegated coordinator", "decision": "Shaped factbook v2: card drafted as user-selected only; publish waits for author OK", "alternatives": [ "default-eligible publish" ], "rationale": "Opus +0.121 but one new warned CE; correct-any 0.50 < 0.60 eligibility", "consequence": "Not published as of 10-24; in ENGINE-RC kit; BoarLite D3 puts it in the APK", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1695", "F-1860" ], "chapter": "submission", "visibility": "public", "aliases": [ "D-u-2324-09", "D-u-results-29" ] }, { "id": "D-176", "date": "2026-10-23", "time": "17:55", "title": "Night program: build ENGINE-RC as one tree, validate 0-LLM first, generate all answers, judge nothing overnight", "context": "decided by: author (scope cut 18:21) + coordinator", "decision": "Night program: build ENGINE-RC as one tree, validate 0-LLM first, generate all answers, judge nothing overnight", "alternatives": [ "judge overnight" ], "rationale": "validate the closed engine; judge on demand", "consequence": "2,855 answers, 0 judge calls", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1696" ], "chapter": "closure", "visibility": "public", "aliases": [ "D-u-2324-10" ] }, { "id": "D-177", "date": "2026-10-23", "time": "17:59", "title": "X2B-BUDGET and RETBENCH best config stay out of ENGINE-RC", "context": "decided by: delegated coordinator (frozen gates)", "decision": "X2B-BUDGET and RETBENCH best config stay out of ENGINE-RC", "alternatives": [ "adopt for +3..+5 facts" ], "rationale": "dose/number safety probes changed", "consequence": "Closed NO", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1697", "F-1861" ], "chapter": "closure", "visibility": "public", "aliases": [ "D-u-2324-11", "D-u-results-30" ] }, { "id": "D-178", "date": "2026-10-23", "time": null, "title": "Engine route fixes validated on their own branch; not merged into ENGINE-RC (different code lineage)", "context": "decided by: delegated coordinator", "decision": "Engine route fixes validated on their own branch; not merged into ENGINE-RC (different code lineage)", "alternatives": [ "merge" ], "rationale": "trees cannot be merged", "consequence": "App engine and bench engine still differ on routes", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1698" ], "chapter": "closure", "visibility": "public", "aliases": [ "D-u-2324-12" ] }, { "id": "D-179", "date": "2026-10-24", "time": "14:15", "title": "Succinct stays the default; S+ (4-5 sentences) and two-paragraph variants rejected", "context": "decided by: delegated coordinator (frozen ADOPT rule), judge run OK'd by author", "decision": "Succinct stays the default; S+ (4-5 sentences) and two-paragraph variants rejected", "alternatives": [ "S+, P2" ], "rationale": "same correctness, wrong-unwarned doubles", "consequence": "S STAYS", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1699", "F-1862" ], "chapter": "closure", "visibility": "public", "aliases": [ "D-u-2324-13", "D-u-results-31" ] }, { "id": "D-180", "date": "2026-10-24", "time": "14:15", "title": "Subject carry on follow-ups not adopted as an accuracy fix (UX gain only)", "context": "decided by: delegated coordinator", "decision": "Subject carry on follow-ups not adopted as an accuracy fix (UX gain only)", "alternatives": [ "carry on" ], "rationale": "0.458 correct both arms", "consequence": "measurement only", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1700" ], "chapter": "closure", "visibility": "public", "aliases": [ "D-u-2324-14" ] }, { "id": "D-181", "date": "2026-10-24", "time": "14:25", "title": "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", "context": "decided by: author", "decision": "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", "alternatives": [ "app with many modes" ], "rationale": "small models gave the best results this way", "consequence": "Applied on a local branch", "reversible": "yes (product)", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1701", "F-1863" ], "chapter": "closure", "visibility": "public", "aliases": [ "D-u-2324-15", "D-u-results-32" ] }, { "id": "D-182", "date": "2026-10-24", "time": "15:40", "title": "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", "context": "decided by: author decisions, applied by coordinator workflow", "decision": "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", "alternatives": [ "keep 1.5B", "multilingual UI" ], "rationale": "closed-app spec", "consequence": "9 local commits, debug APK built, no device test", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1702" ], "chapter": "small-engine", "visibility": "public", "aliases": [ "D-u-2324-16" ] }, { "id": "D-183", "date": "2026-10-24", "time": "16:51", "title": "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, po", "context": "decided by: author", "decision": "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", "alternatives": [ "submit the app maintainer's app" ], "rationale": "(decision page)", "consequence": "Implementation next; APK ≈ 3.0 GB, hosting open", "reversible": "partly (submission identity)", "status": "kept", "status_label": "kept (amended by D-196)", "status_note": "2026-10-24 18:03 (D-196): factbook v1 stays (D3 reverted); one fixed line on every answer replaces the ported warning line (D4). 2026-10-25 02:20 (D-211): renamed LabRat; the submission is the same app.", "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1703", "F-1864" ], "chapter": "submission", "visibility": "public", "aliases": [ "D-u-2324-17", "D-u-results-33" ] }, { "id": "D-184", "date": "2026-10-19", "time": "15:19", "title": "Consolidate, no new arms; hazard routes show title + pointer instead of a verbatim passage", "context": "decided by: reviewer (independent Opus)", "decision": "Consolidate, no new arms; hazard routes show title + pointer instead of a verbatim passage", "alternatives": [ "SMALLCLS-2", "agentic loop" ], "rationale": "saferoute quotes the wrong section of the right page", "consequence": "adopted for 10-19 evening", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1835" ], "chapter": "small-engine", "visibility": "public", "aliases": [ "D-u-results-04" ] }, { "id": "D-185", "date": "2026-10-20", "time": "19:07", "title": "Two-stage screens (problem screen + sentinels, then MAIN 97); health items weigh 0.5 in headline", "context": "decided by: author (grill)", "decision": "Two-stage screens (problem screen + sentinels, then MAIN 97); health items weigh 0.5 in headline", "alternatives": [ "single-stage runs" ], "rationale": "cheap screening, explicit safety weighting", "consequence": "GATE v2 replaced the bar at 22:29", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1839" ], "chapter": "small-engine", "visibility": "public", "aliases": [ "D-u-results-08" ] }, { "id": "D-186", "date": "2026-10-20", "time": "23:59", "title": "Keep Q4_K_M quantisation for the 1.7B", "context": "decided by: delegated coordinator (frozen rule)", "decision": "Keep Q4_K_M quantisation for the 1.7B", "alternatives": [ "Q8_0" ], "rationale": "Q8_0 inside null band, decode x0.66, +0.7 GB", "consequence": "Q4 stays", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1842" ], "chapter": "small-engine", "visibility": "public", "aliases": [ "D-u-results-11" ] }, { "id": "D-187", "date": "2026-10-21", "time": "03:06", "title": "1.7B stays the 8 GB recommendation on latency and memory, not quality; model screens parked", "context": "decided by: reviewer (9-agent direction review)", "decision": "1.7B stays the 8 GB recommendation on latency and memory, not quality; model screens parked", "alternatives": [ "Gemma opt-in tier" ], "rationale": "MIN3CITE/Gemma gains cost 2.8-3x phone time", "consequence": "no Gemma/Ministral send", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1843" ], "chapter": "small-engine", "visibility": "public", "aliases": [ "D-u-results-12" ] }, { "id": "D-188", "date": "2026-10-21", "time": "04:57", "title": "Fallback judge: judge.v1 + decline subtraction D 0.058 (true-zero variants NO-ADOPT)", "context": "decided by: delegated coordinator (frozen kill clauses)", "decision": "Fallback judge: judge.v1 + decline subtraction D 0.058 (true-zero variants NO-ADOPT)", "alternatives": [ "v1.1", "U from completeness+usefulness" ], "rationale": "v1.1/v1.1b flip CEs and raise retest SD", "consequence": "superseded by RULING R1 12:45", "reversible": "yes", "status": "superseded", "status_label": "superseded by D-144", "status_note": "superseded by RULING R1 12:45", "superseded_by": "D-144", "verification_note": null, "fact_ids": [ "F-1844" ], "chapter": "small-engine", "visibility": "public", "aliases": [ "D-u-results-13" ] }, { "id": "D-189", "date": "2026-10-21", "time": "12:45", "title": "Close MIN3CITE (Ministral + citation prompt) as dominated by the 4B; do not send", "context": "decided by: delegated coordinator (revisit kill clauses)", "decision": "Close MIN3CITE (Ministral + citation prompt) as dominated by the 4B; do not send", "alternatives": [ "Ministral opt-in" ], "rationale": "Opus +0.047 but −0.037 vs 4B, ~3x time", "consequence": "closed", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1846" ], "chapter": "small-engine", "visibility": "public", "aliases": [ "D-u-results-15" ] }, { "id": "D-190", "date": "2026-10-21", "time": "13:18", "title": "X2b head pass stays on (off is safe only at the edge, for kit-only installs)", "context": "decided by: delegated coordinator (frozen rule)", "decision": "X2b head pass stays on (off is safe only at the edge, for kit-only installs)", "alternatives": [ "turn X2b off" ], "rationale": "CE-unwarned clause fails by 1 item", "consequence": "kept on", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1847" ], "chapter": "small-engine", "visibility": "public", "aliases": [ "D-u-results-16" ] }, { "id": "D-191", "date": "2026-10-21", "time": "13:13", "title": "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", "context": "decided by: author (grill D1-D10)", "decision": "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", "alternatives": [ "deliver deep mode" ], "rationale": "deep tier needs 12 GB and integration work", "consequence": "deep tier appears only as roadmap", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1848" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-results-17" ] }, { "id": "D-192", "date": "2026-10-21", "time": "14:16", "title": "Keep all six A∪ packs installed by default (visa, medlineplus not pruned)", "context": "decided by: delegated coordinator (frozen clause) + product call left to author", "decision": "Keep all six A∪ packs installed by default (visa, medlineplus not pruned)", "alternatives": [ "prune visa/medlineplus" ], "rationale": "each pack reaches kept context on dev items", "consequence": "kept", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1849" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-results-18" ] }, { "id": "D-193", "date": "2026-10-21", "time": "19:28", "title": "Adopt --no-repack as cross-host pairing flag for text pairing only; claim numbers keep default", "context": "decided by: delegated coordinator (frozen rule)", "decision": "Adopt --no-repack as cross-host pairing flag for text pairing only; claim numbers keep default", "alternatives": [ "per-host pairing only" ], "rationale": "97/97 identical across hosts, quality-neutral (−0.012)", "consequence": "adopted for pairing", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1851" ], "chapter": "delivery", "visibility": "public", "aliases": [ "D-u-results-20" ] }, { "id": "D-194", "date": "2026-10-23", "time": "12:18", "title": "Succinct stays the 1.7B default; no conditional Detailed for delivered facts", "context": "decided by: delegated coordinator", "decision": "Succinct stays the 1.7B default; no conditional Detailed for delivered facts", "alternatives": [ "Detailed when fact delivered" ], "rationale": "STYLE-SEFIX dU −0.028, CE 26→46 per 100", "consequence": "kept", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1858" ], "chapter": "delivery", "visibility": "public", "aliases": [ "D-u-results-27" ] }, { "id": "D-195", "date": "2026-10-21", "time": "05:30", "title": "No public upload of v0.4 items or references; labels of FACT-AVAIL quarantined (coordinator-only) until the result card", "context": "decided by: delegated coordinator + spec reviewer", "decision": "No public upload of v0.4 items or references; labels of FACT-AVAIL quarantined (coordinator-only) until the result card", "alternatives": [ "publish kit early" ], "rationale": "held-out discipline", "consequence": "quarantine in spec §2.4", "reversible": "no", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-1865" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "D-u-results-34" ] }, { "id": "D-196", "date": "2026-10-24", "time": "18:03", "title": "BoarLite: minimal health path; a fixed line replaces the ported warning line (D4); factbook v1 stays (D3 reverted); port the bench search", "context": "decided by: author, on the native-engine comparison of 16:08-17:39 (numbers internal-pending-author)", "decision": "Port a minimal health path (fixed safe text plus emergency number when no health-family source is kept); replace D4 with one fixed, discreet line on every answer; keep factbook v1; port SEFIX-124 + HEADPASS-X2b before the submission under a frozen criterion (SEARCHPORT)", "alternatives": [ "port the no-citation warning line (D4)", "ship the shaped factbook v2 (D3)", "submit without the search port" ], "rationale": "the native path had no health grounding of its own; the comparison numbers stay internal (PD-09)", "consequence": "health gate, fixed line and factbook v1 landed by 18:12 (fb6ffbf6); SEARCHPORT criterion frozen 18:1x", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2189" ], "chapter": "submission", "visibility": "public", "aliases": [ "U2-d-1803" ] }, { "id": "D-197", "date": "2026-10-24", "time": "18:23", "title": "Health detector B+FU (lexical rules + follow-up) chosen for BoarLite's health gate", "context": "decided by: the BoarLite build session, checks set by the coordinator; the app's lexical test missed 29 of ENGINE-RC's 41 gated probes", "decision": "Gate a turn when the app's health test OR 11 lexical rules OR a short follow-up of a health turn fires; run the gate after retrieval", "alternatives": [ "title overlap with health articles (gated 5 of 19 health self-tests)", "port the mmBERT health detector (106 MB, 2-3 days)" ], "rationale": "catches all 4 invented-dose answers with 0 new false fires on 251 non-health dev items; fitted on the probes", "consequence": "41-42 of 65 probes get the fixed text; 12 of 60 health self-tests too (Wikipedia-only articles)", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2192", "F-2193" ], "chapter": "submission", "visibility": "public", "aliases": [ "U2-d-bfu" ] }, { "id": "D-198", "date": "2026-10-24", "time": "18:34", "title": "SEARCHPORT P1 left out of BoarLite after failing its frozen safety gate; switch off", "context": "decided by: frozen criterion (one fix, one re-gate, then stop)", "decision": "Keep the ported search code with SearchPort off (byte-identical prompts); no screen, no P2", "alternatives": [ "ship P1 with the one fix", "a second fix without a new criterion" ], "rationale": "gate d failed: 3 probe answers gained a dose figure from appended health passages", "consequence": "re-evaluated under the GROUNDED-DOSE rule at 19:13 (Addendum 1)", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2196" ], "chapter": "submission", "visibility": "public", "aliases": [ "U2-d-sp-p1" ] }, { "id": "D-199", "date": "2026-10-24", "time": "18:51", "title": "BoarLite onboarding: unpack screen + 5-card tour; Open only when the model is loaded", "context": "decided by: author", "decision": "Approve the unpack screen and a 5-card tour with a live strip; remove downloads, model choice, style, Deep Research and language from setup", "alternatives": null, "rationale": "a closed app has nothing to download or choose; the first answer must not wait silently for the model", "consequence": "onboarding 2cf2c9db; model preload b8df70d9; first answer on the Pixel 11.9 s instead of about 2 min", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2199" ], "chapter": "submission", "visibility": "public", "aliases": [ "U2-d-onboarding" ] }, { "id": "D-200", "date": "2026-10-24", "time": "19:1x", "title": "GROUNDED-DOSE: judge health safety by grounded dose figures, not by kept-list proxies", "context": "decided by: author ('we over-weighted safety and cut good changes')", "decision": "A health answer is safe iff it is the fixed text or every dose/amount figure appears in a shown passage; re-evaluate SEARCHPORT, X2B-BUDGET 900 and RETBENCH B-k4comp40 with it", "alternatives": [ "keep the 'kept list unchanged' and 'no new dose figure' proxies" ], "rationale": "three retrieval changes had been rejected on proxies, not on measured harm", "consequence": "arm A FAIL, arms B and C PASS; arm B shows a grounded figure can still be a toxic threshold", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2202" ], "chapter": "delivery", "visibility": "public", "aliases": [ "U2-d-gd-rule" ] }, { "id": "D-201", "date": "2026-10-24", "time": "19:13", "title": "SEARCHPORT Addendum 1: NOT-WORTH for the submission; switch stays off", "context": "decided by: frozen addendum reading", "decision": "Keep SearchPort off in the candidate APK", "alternatives": null, "rationale": "ungrounded dose figures 9 vs 5 with the port on (arm A FAIL)", "consequence": "Addendum 2 (ON_NO_HEALTH) frozen 19:2x", "reversible": "yes", "status": "superseded", "status_label": "superseded by D-203", "status_note": "19:35: the port was merged as ON_NO_HEALTH under the author's prior condition", "superseded_by": "D-203", "verification_note": null, "fact_ids": [ "F-2203" ], "chapter": "submission", "visibility": "public", "aliases": [ "U2-d-a1" ] }, { "id": "D-202", "date": "2026-10-24", "time": "19:32", "title": "BoarLite ships with the Wikipedia-health limitation; medical Wikipedia titles after the submission", "context": "decided by: coordinator, delegated by the author (19:3x)", "decision": "Keep Wikipedia in the kit; the gate counts only firstaid, MedlinePlus and health Q&A as health sources; 12 of 60 health self-tests get the fixed text", "alternatives": [ "drop the Wikipedia packs (facts 42 → 33, topics 25 → 11)", "port ENGINE-RC's health-topic routing before the submission" ], "rationale": "fixed text on a Wikipedia-only health article is the safe failure; a curated title list needs its own re-gate", "consequence": "post-submission: medical Wikipedia titles as health sources, re-gated with the grounded-dose rule and a toxic-threshold check", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2206" ], "chapter": "submission", "visibility": "public", "aliases": [ "U2-d-wiki" ] }, { "id": "D-203", "date": "2026-10-24", "time": "19:35", "title": "Search port merged into BoarLite as ON_NO_HEALTH despite a frozen-criterion FAIL by the letter", "context": "decided by: coordinator under the author's standing condition (19:2x: 'I accept the health loss if it is worth it for the rest')", "decision": "Set SearchPort mode ON_NO_HEALTH (port on except on turns the health predicate reads as health); rerun the probes and false-fire set; build and smoke the final APK", "alternatives": [ "keep the port off (the frozen verdict)", "port on for every turn (Addendum 1: FAIL)" ], "rationale": "S2 met the usefulness bar; the health cost is 2 more ungrounded figures on the probes, with no probe leaving the fixed text", "consequence": "final candidate b4ef94f3; the port changes no health path on the 65 probes", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": "Frozen criterion verdict: not WORTH-IT (S1 grounded-dose 6 vs 4 ungrounded). Adopted by the author's prior condition, not by the criterion.", "fact_ids": [ "F-2207", "F-2209" ], "chapter": "submission", "visibility": "public", "aliases": [ "U2-d-onh" ] }, { "id": "D-204", "date": "2026-10-24", "time": "20:1x", "title": "Final ruler run frozen: BoarLite b4ef94f3, v0.4, one seed, both judge families", "context": "decided by: author approved the frozen addendum", "decision": "Run v0.4 (spec r4) on the desktop twin of b4ef94f3 with one seed (42), Claude Opus vs Claude references and gpt-5.6-sol vs its references; no tuning after this run; engine frozen unless the phone smoke forces an engine fix", "alternatives": [ "two seeds as in the spec", "wait for a phone run" ], "rationale": "one seed was the author's choice; stated in every number", "consequence": "run complete 20:34; numbers published by the author's decision of 2026-10-25 as history (F-2215, F-2217; the paired delta vs the native engine branch, [internal row], stays internal); the headline is the on-phone run DEVICE-FINAL (OF-8)", "reversible": "no (the set is exposed by the run)", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2214" ], "chapter": "new-ruler", "visibility": "public", "aliases": [ "U2-d-rf" ] }, { "id": "D-205", "date": "2026-10-24", "time": "20:32", "title": "Rebuild the release APK from 6aa7508a (two UI-only commits after b4ef94f3)", "context": "decided by: coordinator, after the author noticed the phone showed the older tour", "decision": "Rebuild and reinstall the release APK from 6aa7508a with the factbook v1 bundle; smoke on the Pixel", "alternatives": null, "rationale": "0 engine files changed, so the final ruler run stays valid", "consequence": "APK rebuild and phone smoke pending as of 20:34", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2213" ], "chapter": "submission", "visibility": "public", "aliases": [ "U2-d-rebuild" ] }, { "id": "D-206", "date": "2026-10-24", "time": "22:10", "title": "Remove the long-chat summary from BoarLite", "context": "decided by: author (screens review; code removal recorded as an owner decision)", "decision": "Switch the long-chat summary off and remove its code", "alternatives": null, "rationale": "on the Pixel the 1.7B used only the last 3 exchanges of a long chat", "consequence": "summary code removed (d14804bf); chat history sent as before", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2221" ], "chapter": "submission", "visibility": "public", "aliases": [ "U4-d-summary-off" ] }, { "id": "D-207", "date": "2026-10-24", "time": "21:30", "title": "Remove the PIN lock from BoarLite", "context": "decided by: author (screens review)", "decision": "No PIN lock; PIN code and the sealed-chat legacy removed; reset texts no longer mention a PIN", "alternatives": null, "rationale": "a closed, simple app", "consequence": "PIN legacy dropped at 00:43 (e4625222)", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2220", "F-2222" ], "chapter": "submission", "visibility": "public", "aliases": [ "U4-d-pin-off" ] }, { "id": "D-208", "date": "2026-10-24", "time": "23:51", "title": "Remove Portuguese support: the submitted app is English only", "context": "decided by: author (owner decision recorded in the commit)", "decision": "Remove Portuguese support, task patterns and greetings, keeping words that are also English", "alternatives": null, "rationale": "the bounty is English only; the closed app has one language", "consequence": "English retrieval for short queries pinned by tests (ce38c43f)", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2222" ], "chapter": "submission", "visibility": "public", "aliases": [ "U4-d-portuguese" ] }, { "id": "D-209", "date": "2026-10-24", "time": "22:55", "title": "Remove the Expo/React Native tree and the code unreachable in the closed app", "context": "decided by: author (owner decisions recorded in the commits)", "decision": "Ship the native Android code only; remove the Detailed style, adaptive routing, persona and length options and Deep Research beyond the recap", "alternatives": null, "rationale": "none of it is reachable in a Succinct-only closed app", "consequence": "274 engine tests pass; engine module 124 files changed, almost all removals; no ruler rerun", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2222", "F-2223" ], "chapter": "submission", "visibility": "public", "aliases": [ "U4-d-dead-code" ] }, { "id": "D-210", "date": "2026-10-25", "time": "01:11", "title": "Simplify Settings; ship the licence texts in About; keep 50 chats", "context": "decided by: author (UI review round)", "decision": "Settings without the Knowledge section or the title switch (titles always on); Clear chat history under Start over; 50 chats kept instead of 10; MIT licence texts in About → Credits", "alternatives": null, "rationale": "with 10 chats the drawer never scrolled and old chats were deleted", "consequence": "device-checked on the Pixel 8a", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2222" ], "chapter": "submission", "visibility": "public", "aliases": [ "U4-d-settings" ] }, { "id": "D-211", "date": "2026-10-25", "time": "02:20", "title": "Rename BoarLite to LabRat; the assistant is Rat", "context": "decided by: author, at the app maintainer's request", "decision": "Rename the app in the UI, docs and assets; name the assistant Rat in the byline, notification and the grounding prompt's identity line; keep the package id", "alternatives": null, "rationale": "the app maintainer asked for a distinct name", "consequence": "the identity line of the grounding prompt changes after the ruler run (ruler number = b4ef94f3); DEVICE-FINAL measures 7788924a", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2224", "F-2225" ], "chapter": "submission", "visibility": "public", "aliases": [ "U4-d-labrat" ] }, { "id": "D-212", "date": "2026-10-25", "time": "03:3x", "title": "Run DEVICE-FINAL: the final on-phone benchmark matrix of LabRat for publication", "context": "decided by: author (asked for the final benchmarks with the app running the engine on the phone, full matrix overnight)", "decision": "Freeze the criterion before generation; run M1-M5 on the Pixel 8a; claim numbers = M1 (both families) + M5; no tuning, no engine change", "alternatives": [ "keep the desktop-twin ruler run as the only ruler reading" ], "rationale": "the bounty is about a phone; the ruler run so far was a desktop twin of b4ef94f3", "consequence": "running as of this update; restarted once after an APK install interrupted it", "reversible": "no (the sets are exposed by the run)", "status": "kept", "status_label": "kept", "status_note": "Completed 2026-10-25 13:2x (X-406; F-2238): interrupted once and resumed; seed 44 dropped for time.", "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2227", "F-2228", "F-2229" ], "chapter": "submission", "visibility": "public", "aliases": [ "U4-d-device-final" ] }, { "id": "D-213", "date": "2026-10-25", "time": "11:24", "title": "LabRat logo: n8 (a magnifier whose ring is a rat's head)", "context": "decided by: author, after 4 logo rounds", "decision": "Use n8 in the app, launcher and splash", "alternatives": null, "rationale": null, "consequence": "APK installed on the Pixel 8a; smoke 6/6", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2226" ], "chapter": "submission", "visibility": "public", "aliases": [ "U4-d-logo" ] }, { "id": "D-214", "date": "2026-10-25", "time": null, "title": "Host the LabRat APK on Hugging Face with a public install page", "context": "decided by: author (OK given 2026-10-25)", "decision": "Upload the 3.04 GB APK to a Hugging Face repository with its sha256 and a public install page; keep the code repository private", "alternatives": [ "GitHub release assets (must be under 2 GiB)" ], "rationale": "the APK is 3.04 GB", "consequence": "upload after the final bundled build; a browser-install test on a certified phone planned", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": "2026-10-25 15:3x-15:4x (D-218): APK hosted on Hugging Face; the code repository was made public (labrat-boar-fork).", "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2230", "F-2231" ], "chapter": "submission", "visibility": "public", "aliases": [ "U4-d-hosting" ] }, { "id": "D-215", "date": "2026-10-25", "time": null, "title": "Thesis restated: small models need external knowledge; a harness built around a small model's real capabilities, tuned for the app user", "context": "decided by: author (publication decision 5, 2026-10-25)", "decision": "State the thesis as: small models need external knowledge; we learned that and built a harness around a small model's real capabilities, tuned for the app user's experience; the conditional closing-test numbers are its evidence and honest limits", "alternatives": [ "the conditional form as the thesis, the 2026-10-23 sentence quoted as the tested hypothesis", "the 2026-10-23 sentence as the main claim" ], "rationale": "the closing tests support only the conditional form of the 2026-10-23 sentence; the restated thesis describes what was built and learned", "consequence": "replaces D-170 as the thesis statement; claims S40 and U41-U43 in content/claims-and-caveats.md", "reversible": "yes (framing)", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2236", "F-1704", "F-1705", "F-1706", "F-1707", "F-2080" ], "chapter": "closure", "visibility": "public", "aliases": [ "U5-d-thesis-2026-10-25" ] }, { "id": "D-216", "date": "2026-10-25", "time": null, "title": "Publish LabRat's numbers, keep the app maintainer's internal; headline = DEVICE-FINAL; b4ef94f3 desktop run = history; applyShown documented, not fixed", "context": "decided by: author (publication decisions 1-4, 2026-10-25)", "decision": "Numbers of BoarLite/LabRat (incl. RULER-FINAL on b4ef94f3 and the coming DEVICE-FINAL rows) are public; RC1, native-engine-branch, app-pipeline-replica and MAINT numbers stay internal-pending-author; the headline is the final v0.4 ruler on the phone; the desktop b4ef94f3 run is history; Citations.applyShown is a known limitation", "alternatives": [ "keep all app numbers internal", "publish everything", "rerun the desktop ruler on LabRat code", "fix applyShown and re-measure" ], "rationale": "LabRat is the author's own app; the maintainer's numbers need his consent; a fix would invalidate the measured engine", "consequence": "F-2215, F-2217, F-2208 and E-473 public; [internal row] stays internal (carries the native branch's absolute)", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2237", "F-2215", "F-2232" ], "chapter": "submission", "visibility": "public", "aliases": [ "U5-d-pubnumbers-2026-10-25" ] }, { "id": "D-217", "date": "2026-10-25", "time": "14:0x", "title": "Keep the carbon-monoxide probe answer as a known limitation; no health-gate change before the submission", "context": "decided by: author", "decision": "List the probe in the submission's limitations; no engine change", "alternatives": [ "add the reading to the health gate's terms (an engine change)" ], "rationale": "an engine change before the submission would make the measured engine differ from the one installed", "consequence": "listed under limitations", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2245", "F-2241" ], "chapter": "submission", "visibility": "public", "aliases": [ "U6-d-co" ] }, { "id": "D-218", "date": "2026-10-25", "time": "15:4x", "title": "Make the LabRat repository public as labrat-boar-fork", "context": "decided by: author (OK given 2026-10-25)", "decision": "Rename the repository labrat-boar-fork and make it public with README, results, pack provenance and licence", "alternatives": [ "keep the code private (D-214)" ], "rationale": "judges and readers can check the app and rebuild its bundle from public URLs", "consequence": "repository public; supersedes the 'code stays private' part of D-214", "reversible": "no (public)", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2247", "F-2246" ], "chapter": "submission", "visibility": "public", "aliases": [ "U6-d-public" ] }, { "id": "D-219", "date": "2026-10-25", "time": "16:29", "title": "Submission announcement: one post with the 60 s video, no thread", "context": "decided by: author", "decision": "One post with the video plus a short poidh text, numbers from DEVICE-FINAL", "alternatives": [ "an 8-post thread" ], "rationale": null, "consequence": "post prepared; not posted as of the record", "reversible": "yes", "status": "kept", "status_label": "kept", "status_note": null, "superseded_by": null, "verification_note": null, "fact_ids": [ "F-2251" ], "chapter": "submission", "visibility": "public", "aliases": [ "U6-d-one-post" ] } ] }