Seven readers, one finding

# Pre-Publication Panel Report — 2026-07-04

**Method.** Seven independent expert readers, run on the book AS A READER RECEIVES IT (clean body-prose builds; no frontmatter, no revision logs, no canon, no critic notes; readers blind to each other): one close reader per act (the Act V reader also read Act IV for context), one whole-book structural auditor on a spine of all 34 verbatim chapter openings+closings in sequence, one voice specialist running the stripped-tag test on six dialogue chapters (ch 8, 11, 18, 25, 30, 32). Anchored scale: 6 = competent midlist; 7 = strong literary debut; 8 = prize-longlist; 9 = prize-shortlist; 10 = enduring masterpiece.

**Aggregate verdict: FIXABLE-ISSUES (6 of 7 readers); voices READY (1 of 7); none NOT-READY.**
The book reads at ~8 overall, with 9s at the load-bearing points and 7s in pacing across all five acts.

## Scores

| Reader | Scores | Verdict |
|---|---|---|
| Act I (ch1–7) | craft 8, narrator 8, voices 8, pacing 7, ambiguity 8, consistency 8 | FIXABLE-ISSUES |
| Act II (ch8–15) | craft 8, narrator 9, voices 8, pacing 7, ambiguity 8, consistency 9 | FIXABLE-ISSUES |
| Act III (ch16–22) | craft 9, narrator 9, voices 8, pacing 7, ambiguity 9, consistency 9 | FIXABLE-ISSUES |
| Act IV (ch23–30) | craft 9, narrator 9, voices 8, pacing 7, revelation 9, consistency 7 | FIXABLE-ISSUES |
| Act V (ch31–34) | craft 8, narrator 9, force 8, pacing 7, ending 8, consistency 8 | FIXABLE-ISSUES |
| Structural audit (spine) | openings 7, closings 6, escalation 6, quiet 7, frame 7 | FIXABLE-ISSUES |
| Voice exam | okonkwo 9, rahmani 8, tanwei 7, pope 8, narrator-bleed 7, mouthpiece-risk 7 | READY |

## Convergent thesis (all readers, independently)

The book's own method — pattern-detection — trains its target reader to catch the AUTHOR's patterns, and the device economy is too narrow: roughly six close-mechanisms serving thirty-four chapters. A reader of the whole book, unlike the writer (or auditor) of any single chapter, feels the repeated devices arrive on schedule and "begins to read the rhythm instead of the story."

1. **CONFIRMED — Ch 28 misattributes the book's hinge line.** […] Ch 28's own frontmatter (F174) carries the SAME misattribution — root cause of the audit miss: the error is in both layers, so per-chapter checks agreed with themselves. […]

## Pacing: why five prior passes did not solve it (history reconstructed 2026-07-04)

Prior attempts, in order: AI-tell reduction (bce871e, all chapters, phrase-level); cast re-voicing (495b4cb); v3 "de-templating" (ae7563f — close-mechanism variation enforced PROSPECTIVELY via per-chapter "do not reuse" craft-watch lists); dry-wit/let-it-breathe (0d1c94b); Pass A tic-thinning (6b79bc7, 10 chapters); B1 stencil dedup (earlier, never re-run); Pass B B2+B3 (77a5b50 — intra-chapter one-statement-per-idea + 3 structural interventions, gated by per-chapter blind A/B).

Why the panel's pacing findings survived all of them:

1. **Wrong altitude.** Every pass attacked repetition WITHIN a chapter (B2, tics) or at PHRASE level across chapters (B1 stencils, AI-tells). No pass ever attacked DEVICE-level repetition across chapters — the close-mechanisms, the Rahmani exit, the chapter architecture. That is exactly the altitude the panel's spine audit reads at.
2. **The craft-watch lists were enumerative and ran one device behind.** Each chapter's frontmatter forbade reusing SPECIFIC named closes (the cat-at-the-bell, the corner window, the folded coat…) — so each new close was "fresh" against the list — but "He named nothing," the question-left-standing exit, and the not-opened door were never put ON the list, so they were reinvented chapter after chapter while formally passing.
3. **B2's governing rule made the fix impossible by construction.** "Cuts are subtraction, never replacement with fresh coins" is the right rule against purple restatement, but sameness of landing gear cannot be CUT away — you cannot subtract your way from six devices to twelve. The only permitted tool was the one tool that cannot solve this problem.
4. **The QA bar itself was per-chapter.** A blind juror reading one chapter cannot see that its close also landed three chapters running; per-chapter juries structurally confirm cross-chapter repetition as "in voice" (it is!). The 2026-07-04 spine audit is the first gate that ever read the chapter edges in sequence.
5. **Mass was never the problem.** Deep cuts failed jury review every time they were tried (ch 20 restored 85/15; ch 28 restored 80/20; ch 14 and ch 16 70/30); the panel likewise scored the quiet chapters as earned and praised the stillness. The book is not too slow. It lands too many chapters on the same six mechanisms.