[tweak] Coherence Theater — resolve clipping, caption-safe layout, and final cue coverage #106

Which book?

Coherence Theater — ssrn-aei

Chapter (optional)

All chapters (1–5)

What should change?

This is a release-blocking accessibility and compositing review of the deployed book:
https://orly.brett-lamy.workers.dev/?bundle=ssrn-aei

Please make the smallest coherent scene/source changes that resolve the following, then regenerate affected narration/assets and verify the complete book.

Cross-cutting requirement

Treat the fixed upper-left title/citation strip and the lower caption region as protected areas. With closed captions enabled, the lower approximately 25% of the 1280×720 stage is not available for active, must-read content. Check both one- and two-line captions. Do not solve an active explanatory-state collision merely by letting its subject be clipped or obscured during a camera move.

Required corrections

1. Chapter 1, ~0:32 — title/citation collision. Keep the controlled evaluation — stable framing card out of the fixed upper-left metadata strip. Reframe it below the strip, or deliberately set aside/fade the metadata during the move.
2. Chapter 1, ~1:28; Chapter 2, ~1:25 — caption-safe layout. Move the active lower-right figures out of the caption band. In Chapter 2 this includes the trust-chain graphic.
3. Chapter 2, ~0:15 — grid/list collision. Keep the explanatory property list in its left field and bring the blue 60-person preference-falsification grid into a distinct right-side field, with a readable gutter.
4. Chapter 3, ~0:35–0:45 — right-edge clipping. Fit/reflow the active 6.2 contradictory mandate collapse module before its heading, pink resolution text, and arrow become the focal explanation.
5. Chapter 3, ~0:49 onward — left-edge clipping. Reset/reposition the authority-laundering cascade so its heading, source line, node 1, and first connecting edge all have a normal left safe margin when the cascade is primary.
6. Chapter 4, ~0:05 — left-edge clipping. Adjust the initial camera target/scale or shift the 6.4 narrative capture / performative alignment figure right. At the early 1.3× zoom its current left edge is approximately −10px, cropping the active title/card.
7. Chapter 5, ~0:54 — caption-safe layout. Move the shelf's answers, in this paper's vocabulary: and all three active bullets above the caption-safe boundary (or restage it compactly in clear upper/right space).
8. Chapter 5, final ~1.6s — caption coverage. The final caption disappears before narration ends. The current manifest has final cue 88.538 and duration 103.561; extend that final cue through narration end or add the missing final cue. Validate against delivered audio, not just the scene's authored hold.

round 0 — equal thirds in Chapter 4 is explicitly not part of this request.

Acceptance criteria

• No active/focal text, card, diagram, or sequence is clipped by a viewport edge at the named moments.
• No active/must-read scene content overlaps the title/citation strip or caption overlay when captions are enabled.
• Chapter 5 captions remain present through the end of the actual MP3 narration.
• Preserve the book's existing factual content and all unaffected chapter beats.
• Run npx tsc --noEmit, npm run build, and node generator/verify-book.mjs --slug ssrn-aei.
• Add a regression-capable safe-area/caption check to the book QA path if feasible; otherwise attach caption-enabled screenshots at the named markers and leave a follow-up note for that generic gate.

Model

Opus 4.8 (claude-opus-4-8)