qa(scores): record v1.0.4-rc2 ledger row (post-Sprint-1 re-measure @c92a393)

qa/scores_ledger.md

@@ -6,10 +6,11 @@
 
 > **Ruler** = `scoring_config_version` (a content hash of the rubric + schema + gate files). Rows under DIFFERENT Ruler values are **NOT directly comparable as a quality trend** — the ruler changed (a rubric recalibration or a new gate moves the number with no change in play quality). Use `python3 qa/scores_db.py --compare` for a ruler-fenced trend; comparing across rulers requires re-scoring an archived transcript under the current ruler. **RC** = the release candidate a run scored (e.g. `v1.0.4-rc1`).
 
-> Rows: **65** · rendered 2026-06-09T18:10:34+00:00
+> Rows: **66** · rendered 2026-06-09T23:47:29+00:00
 
 | Run | When | SHA | Build date | Surface | DM model | Actor model | Scorer | Ruler | RC | Methodology | Story | Mech | AngryDM | Behav | Sat | RRI | Crit | Img% | Pass | Source | Notes |
 |---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
+| v1.0.4-rc2-c92a393 | 2026-06-09T23:47:29+00:00 | c92a393 | 2026-06-09 | GUI-built-app | opus | opus | claude | sc_df34ecd02b4f | v1.0.4-rc2 | 11-gate RRI rc2: VM part-B 5-persona @c92a393 (post 7 Sprint-1 reliability fixes); Mac part-A SKIPPED (disk 3.9Gi) → native_gate=no-handoff gap | 4 | 3 | 3 | GREEN | 5.8 | 2.7 | 1 | 0% | FAIL | VM:/root/worldos-qa/results/RRI-5canon.json @c92a393 | POST-Sprint-1-reliability (7 fixes). PRODUCT IMPROVED vs rc1: criticals 4->1 (the #735 active-PC keystone validated — gone from the 4 personas that had it), sat 4.8->5.8, behavioral GREEN, 4/5 personas clean (adv/narr/vet sat=7,0-crit,no-giveup). RRI 3.6->2.7 is DEPRESSED by 2 non-product effects: (1) -1 native_gate = Mac handoff SKIPPED (disk), not a regression — with handoff ~3.7; (2) -1 no_give_up = newbie's mid-stream STALL (beat streamed partially then froze mid-word, controls locked, no retry/cancel = the #623 DEEPER recovery layer; the heartbeat fixed perceived-hang but not a genuinely-stalled beat). newbie's 1 crit IS that same stall. mech 3.0 = one-duo social-sampling (engine-duo ~3.8). ui_audit FAIL (rc=1, separate from #306/#744 title-bar which is fixed). Same ruler sc_5ac7a1d9103c as rc1 (apples-to-apples). NEXT DOMINO: fix newbie mid-stream-stall recovery -> flips no_give_up + zero_critical. |
 | v1.0.4-rc1-fa97b34 | 2026-06-09T18:10:34+00:00 | fa97b34 | 2026-06-09 | GUI-built-app | opus | opus | claude | sc_5ac7a1d9103c | v1.0.4-rc1 | 11-gate RRI RC1: Mac part-A handoff 100/100 + VM part-B 5-persona sweep (lean-ON); rollup via fixed release_readiness.py (#728) | 3.9 | 3.5 | 3.2 | GREEN | 4.8 | 3.6 | 4 | 0% | FAIL | /Volumes/LEXAR/Codex/worldos-qa-results-fa97b34/RRI-full.json | RC1 honest 11-gate RRI @fa97b34 (post #719/#720). PASS: native_gate(signal; Mac handoff 100/100), arc, no_give_up, behavioral GREEN, palette. FAIL: cross_persona_sat 4.8<7, zero_critical(4), story 3.9<4.3, mech 3.2<4.5(one-duo sampling, understated), ui_audit FAIL(axe). native_gate withheld credit pending --support-preflight (split VM+Mac contract) — handoff itself is valid 100/100. image_render 0% = VM-gateway gap (Mac/gateway lane measures it). vs Tuesday 3.6: no_give_up RED->PASS (input-lock give-up gone), sat 4.4->4.8, behavioral GREEN. Caught+fixed build_shas rollup regression (#728) that had crashed EVERY release_readiness.py rollup since #723. |
 | reorient-leanON-e2c0da4 | 2026-06-09T13:53:41+00:00 | e2c0da4 | 2026-06-09 | engine-duo | opus | opus | claude | sc_52a5cbd26775 |  | 3-lens duo 8-beat lean-ON (CLAWDND_LEAN_BEATS); reorient post-fix baseline | 4.1 | 3.7 | 3.4 | GREEN |  |  |  |  |  | /Volumes/LEXAR/Codex/session-notes/2026-06-05/worldos-release-grind/reorient-leanON-duo-e2c0da4.log | First ruler-stamped baseline (#722). Lean-ON Opus healthy → confirms the lean-ON dip was transient (patched lean re-ground bug). Lens files identical e2c0da4↔52e74e2; only release_readiness.py changed (#723), which does NOT affect engine-duo lens scoring. |
 | handoff-20260607-da05101-current-main-clean | 2026-06-07 | da05101 |  | GUI-built-app | scripted+codex | scripted |  |  |  | hybrid handoff gate (web smoke 5 + built smoke 5 + built Codex 1) |  |  |  |  |  |  |  |  | PASS | /Volumes/LEXAR/Codex/worldos-agent-grade-app-testability/handoff-20260607-da05101-current-main-clean/ | handoff_score=100/100 on current origin/main da05101; clean same-disk worktree, dirty=false, release_verdict=false. Supersedes 9545383 as GUI velocity proof. Full #466 five-persona RRI still required. No quality lens. |