What we shipped this week

Build 4442: Surprise me can no longer roll a "Male" vocal over a soprano seed — the Vocal now matches the seed's voice.

chore: auto-refresh git snapshot [skip snapshot]

Build 4441: the 2026-10-10 Deep Audit's follow-ups are on the backlog — 5 new rows, 4 folded into existing ones, and the audit's own mistake recorded.

chore: auto-refresh git snapshot [skip snapshot]

Build 4440: Deep Audit 2026-10-10 filed; the sitemap parses again, the Crucible stops claiming "zero stored", and the pricing and forge pages describe the pipeline that actually runs.

chore: auto-refresh git snapshot [skip snapshot]

Build 4439: album and Song-Novel track forges now show up in cost telemetry — one row per track.

chore: auto-refresh git snapshot [skip snapshot]

Build 4438: the admin cover batch can no longer overwrite a good cover with fal's black safety frame; today's builds recorded in the backlog.

chore: auto-refresh git snapshot [skip snapshot]

Build 4437: the Song-Novel story pass moves to Claude Opus 5.5 — it won the pre-registered blind test 6-0.

Build 4436: the Song-Novel story pass gets its own model task, and its Opus-vs-Sonnet test is pre-registered.

chore: auto-refresh git snapshot [skip snapshot]

Build 4435: AI cover art no longer saves a black square when fal's safety filter trips — it retries with a people-free prompt, then falls back honestly.

chore: auto-refresh git snapshot [skip snapshot]

Build 4434: the Tournament Forge's extra drafts read draft 1's prompt cache instead of each re-writing it (~$0.40 off every song).

chore: auto-refresh git snapshot [skip snapshot]

Build 4433: Forge v5 goes live — every single forge is now the Tournament Forge (3 drafts, blind debiased judging, legibility check, repair only on a named wound).

chore: auto-refresh git snapshot [skip snapshot]

Build 4432: Forge v5, the Tournament Forge, is built dark and its A/B is pre-registered.

Build 4431: a rejected repair no longer leaves its lyrics in the saved song; the dead cold-reader call is retired.

chore: auto-refresh git snapshot [skip snapshot]

Build 4430 (Master Punch List): the gauntlet writer moves to Claude Opus 5.5; refine measured worse on Opus and stays on Sonnet 4.6.

chore: auto-refresh git snapshot [skip snapshot]

Build 4429 (Master Punch List): the song writer moves to Claude Opus 5.5 at effort medium — measured first, and only the forge core moves.

chore: auto-refresh git snapshot [skip snapshot]

chore: auto-refresh git snapshot [skip snapshot]

chore: auto-refresh git snapshot [skip snapshot]

Build 4428 (Master Punch List): make the request layer Claude-5.x-safe — per-model thinking, max_tokens headroom, thinking-free responses, refusal fallback — and arm the Forge Lab to measure Opus 5.5 on the production prompt. No production route changes model.

Build 4427 (HOTFIX): the forge still 400'd after B4426 — the effort strip never reached the wire. Fixed in the fetch shim and proven against the real API.

Build 4426 (HOTFIX): forge down in production — roll the model routes back to the last-known-good 4.x generation.

Build 4425 follow-up (Master Punch List): production RLS verification recorded; the auto-apply migration pipeline found blocked since ~2026-06-06 — 45-row ledger reconciliation prepared. Docs-only; folds into the next deploy.

chore: auto-refresh git snapshot [skip snapshot]

Build 4425 (Master Punch List): close the two database P0s — scope 12 "service role" RLS policies to service_role, enable RLS on 4 unprotected tables, and stop signed-in users rewriting their own tier/credits.

Build 4424 — docs: refresh AUTO-STATE after the cadence sweep

Build 4424 (cadence) — Bet Review #4, External Audit brief #3, and a scoped Deep Pass; all five cadences green

Build 4423 (cadence) — Quality Council + Trust Decay Audit #11, and the seven surfaces B4422 silently falsified

Build 4422 — docs: point AUTO-STATE at a real commit

Build 4422 — migrate to Claude 5.x generation (Opus 5.5 / Sonnet 5.5 / Haiku 5.5)