[amaybaum]

This is a manuscript-consistency change, kept separate from the Introduction paper (#918) as the owner directed on 2026-10-26. The head is b862df75. It is independently accepted, with the owner's decisions applied and the Theorem A.8 restriction applied and accepted. It is up to date with main 687e885b, CI is green (run 40328483098), and GitHub reports the merge as clean. Merge needs separate owner approval of this exact head.

Review trail

The records are in verification/audits/manuscript/bh-consistency-20261026/.

• FINDINGS.md: the read-only investigation of each site against its support at 687e885b.
• REVIEW.md: its independent review. It confirmed 9 errors and 1 consistent site, reclassified Thm A.9, found the Thm A.7 gap, and listed the missed sites.
• DIFF-REVIEW.md: ACCEPT WITH CHANGES (R1–R4) at 4c0d705e.
• DIFF-RECHECK.md: ACCEPT WITH NITS at 78705611; the nits are applied in b61cbaa9.
• A8-REVIEW.md: the review of the Thm A.8 restriction at 2d13792d. Its verdict was BLOCK, narrowly, on the physical interpretation; fixes A1–A3 are applied in ca159cc9.
• A8-RECHECK.md: ACCEPT at ca159cc9. The owner's three questions are all answered yes: mathematical validity, physical interpretation, and consistency.
• CHANGE-NOTE.md: the site tables, the owner's decisions verbatim (§7), the A.8 restriction (§7–§8), build and check results, and re-grep counts.
• probes/: the exact checks.

What changes

• Conservation versus recovery. Bijectivity gives conservation in the substratum. Recovery from the radiation is a separate, conditional question (ch07, ch18, ch19, README, glossary, appendix-a).
• GR:811. "τ_evap ≪ τ_B^D" is false: about 10⁶² at GR's own 30 M☉ example.
  • Thm A.7 is stated on t ≪ τ_B^D for V∪B, with the error term Thm A.5 step (i) supplies.
  • H-decouple names the closure Thms A.8–A.9 assume: B∪R is a closed system in a pure joint state, on which φ acts as a bijection of its configuration space, with no coupling to D or V∖R, over whole cycles.
• Thm A.8, restricted.
  • Scope. It holds where Page's value is justified: the whole configuration space C_B×C_R is one energy shell, spanning H_B⊗H_R, with H-scramble over its pure states.
  • Physical meaning. The text says plainly that this makes the energy constant on every configuration of B∪R, so energy conservation constrains nothing. It does not follow from conserving a non-constant Hamiltonian, and an evaporating black hole whose radiation carries energy is the open proper-shell case.
  • Proper shells. On a proper shell the mean differs in general: on the diagonal shell it is ½ against Page's 1/3 at m = 2. A.8 and A.9 on proper shells are open.
• Thm A.9. It is restated for full-cycle time averages at a fixed bipartition, on A.8's shell. The single-trajectory curve is open.
• H-scramble. It fails for every bijection from a definite initial configuration with an aligned bipartition. Which remedy the substratum supplies is left open.
• GR:869, :829, :605, :611, :717; ch07; ch09:127; README:64.
  • A.8 requires H-scramble, and the concentration result is Hayden–Leung–Winter's [29, 61].
  • Every "conditional on H-scramble and H-decouple" carries the full-product-space scope.
  • The undefined "horizon-factorization hypotheses" is removed.
• Size of corrections, R12, falsifier.
  • Page's mean has a correction of about ½ nat at equal dimensions.
  • The Page time is a mass-independent fraction only for a mass-independent emission coefficient.
  • The empty falsifier at ch07:161 is removed.
• Bibliography. Page 1993 and Popescu–Short–Winter 2006 are added to the book's bibliography.
• Generated files.
  • GR is 87 pp and the book 554 pp, with no dropped glyphs.
  • The A.7, A.8 and A.9 ledger fingerprints are updated; all stay GAP.
  • dependency_label_check knows the condition name.

Owner decisions (2026-10-26, all applied)

Decision	Owner's recommendation, applied
O-1	Retain the named H-decouple condition
A.9	State the time-averaged result, not pointwise typicality
O-2	Leave the scrambling mechanism open
O-3	Distinguish conservation from observable recovery
O-4	Remove the empty falsifier
O-6	Use "Preserved; Page curve conditional"
A.8	Restrict to the domain where the Page-entropy expression is justified (applied; the energy-shell version is open)

Merge gates

• Independent acceptance (A8-RECHECK at ca159cc9; b862df75 adds only that record)
• Up to date with main 687e885b; mergeable state clean
• CI green on the exact head b862df75 (run 40328483098)
• Owner merge approval of b862df75

🤖 Generated with Claude Code

https://claude.ai/code/session_013T6pfKuv5mi3MXtKjH2tZb

[amaybaum]

Merge record

The owner approved this merge on 2026-10-26 at exact head b862df75.

• Pre-merge checks:
  • the PR head was b862df758428aa6e0695c0dc222ab221b4e0474e;
  • main was 687e885b1cfd24d6f17eee439101afc8acf07015, and the head contains it;
  • the change set was 31 files across the 6 reviewed commits, the same as reviewed.
• Merge commit: 17b830bcd0ecc36e10eff5b9eeca03bdd7198d86, a normal merge commit.
  • Parents: 687e885b (main) and b862df75 (the reviewed head).
  • Tree: 7094a5c5e2f4bc693d59d0ead4cad14d1a2113f0, identical to the reviewed head's tree.
  • The reviewed head is reachable from main.
• CI: green on the PR head b862df75 (run 40328483098) and on the new main 17b830bc (push run 40330683786).
• Out of scope: Thm A.8 on a proper energy shell is recorded as open in GR App. A. Follow-up corrections on the lines this PR edited are in Corpus consistency: horizon entropy, cosmological constant, prediction counts, Cabibbo, echoes #937.

---

Generated by Claude Code