[amaybaum]

Draft control plane of native V3 round K1-RDY-1 (AGENTS.md §A.39). WIP / UNVERIFIED. Not designated F. Nothing is frozen, enumerated or executed.

Planning Issue: #1031. Research record and supersession drafts: #1114 (stays open as the research record, OD-7).

Base and shape

• D = 4b4b437cfff9dcf32c882251d1a4c3445ca9e8be (main, the merge of evidence(#1031 K1-RDY): land PROOFS.md and record the #1119 review (OD-1, OD-2) (WIP / UNVERIFIED) #1122).
• Round K1-RDY-1, non-sealing. Record directory verification/programmes/oi-qm/reconstruction/round-k1-rdy-1/. Receipt path verification/receipts/K1-RDY-1.json.
• Head d9b3de012e393262adb20f91ad49627bad48de43. Five commits from D. Each has a single parent and changes only the round's control-plane files (preregistration.md, amendments/A1–A11):
  1. 6f26bb1b copies proposed-k1-rdy-1/ from research(#1031 K1-RDY): recovery of the accepted preregistration (WIP / UNVERIFIED; not frozen) #1114 at 66bee341 byte for byte.
  2. dfe80bc1 updates the drafts where main or the owner record had moved.
  3. 52ae69fb applies review c.6656052333 under K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6656166518.
  4. 9fa37d0a applies F1/F2 of delta review c.6656432806 under K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6656645169.
  5. d9b3de01 adds the third A10 recording item under K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6657384912 (Option (b)). The hold at 9fa37d0a was released for this one commit.
• Governed paths: only the record directory and the receipt (record AM). Execution is paper-only (OD-4). No execution entry is declared.

Verification at 9fa37d0a. The verify.yml dispatch approved in #1031 c.6657308064 ran at 9fa37d0a. It is evidence for that historical head only, not the final verification of the revised candidate. The owner, c.6657384912: "it must not be treated as the final verification of the revised candidate F".

Authorization

• Drafting: owner, K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6655647525 (10:27:34Z). "Drafting only. Designation of F, freeze, workflow dispatch and scientific execution each require separate approval."
• OD-1 to OD-7: research(#1031 K1-RDY): recovery of the accepted preregistration (WIP / UNVERIFIED; not frozen) #1114 c.6653656756 (08:44:44Z), quoted in §12.
• Later rulings, each quoted where it applies:
  • R1–R5, K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6656166518 (10:54:11Z);
  • F1/F2, broad reading, c.6656645169 (11:18:22Z), in A5;
  • A10 KEEP, c.6657308064 (11:53:35Z), in A10;
  • A10 Option (b), c.6657384912 (11:57:45Z), in A10.
• Earlier rulings: K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6651204512 (enumeration before F not authorized) and c.6650824861.

Protected text

head -n 313 verification/programmes/oi-qm/reconstruction/round-k1-rdy-1/preregistration.md | sha256sum gives
a93f3f33f0c5a6e8b2467312d184fa2d2e915296490948938c4dd7832c64bb63. Checked at all five commits.

Revision d9b3de01: fix → location (paths under round-k1-rdy-1/amendments/)

Fix	Where
Each derivation records three items: (1) its author or producing session; (2) whether the second writer had access to the first derivation; (3) any relevant dependency or shared intermediate result, recording "none" explicitly where there is none	A10 lines 16–20
Shared permitted inputs (the frozen rule, the declared hypotheses Λ, the row specification) are distinguished from intermediate results. Access alone does not establish dependence. A second derivation that relies on the first's intermediate results fails A10's first condition, so the row ends UNDECIDED(G6). The records are procedural and add no gate or token	A10 lines 22–27
The owner's A10 KEEP (c.6657308064, in part) and Option (b) (c.6657384912) are quoted with author and time	A10 lines 34–48
Heads re-read at 2026-11-05T11:58:40Z; all six unchanged	A4 line 17
Trace rows for c.6657384912, c.6657308064 and delta review c.6656892864	A7 lines 46–48

A10's first condition ("takes none of its intermediate results as given") is unchanged and consistent with the new text. No hypothesis, gate, token, conclusion or protected line changed. Earlier revisions are described in the commit messages of 9fa37d0a, 52ae69fb and dfe80bc1.

Checks at d9b3de01

• V3 control plane. The verifier's own control_plane and check_t1 on (D, head) return no codes. 12 control-plane files. Governed-path digest unchanged: c7ecedf33fdab25da74d9ac29e96c13621445d05caafb30d651e869c01300d1d. --self-test OK. No receipt was built.
• Passed: artifact_placement_check.py, duplicate_check.py, staleness_check.py (13 matched), build_migration_manifest.py --check, and SC-42 (PASS, 0 added-text hits). Added text has no local paths and no closing keywords.
• Not run locally. --corpus, legacy_records_check.py and citation_roles_check.py need full history and are undecidable in the shallow local clone. They are left to the delta review or CI.
• Workflows. No push trigger matches this branch or its paths. Only the ordinary verify.yml pull-request run starts.

Remaining before F (each needs separate approval)

1. An independent delta review of d9b3de01 at its exact head.
2. Owner approval of a new verify.yml dispatch on that exact head, with the branch held there.
3. The owner's designation of F.

Not authorized: F designation, freeze, workflow dispatch, enumeration, execution, merge.

🤖 Generated with Claude Code

https://claude.ai/code/session_011zDyUSyEszk3JJx83uXKKq

[amaybaum]

Independent review of the K1-RDY-1 control plane at dfe80bc16718f61137a7964d5f6b4dc9e6fb0152 (pre-F draft)

This is a read-only review, model-authored, by an agent that did not write the PR. I worked from a full, non-shallow clone with no alternates. I made no commits, pushes or dispatches, and I ran no enumeration or model computation.

Verdict: ACCEPT WITH CHANGES. The control plane conforms to §A.39 before F. The protected text is intact, and the owner records are quoted faithfully. Four items need fixing before the owner considers F (R1–R4). R5 is an owner decision that should be made before F.

1. Lifecycle (§A.39, §A.43 item 1, architecture S2/G9/K1/K2/G6/G10/S7)

• D = 4b4b437c is current main and an ancestor of the head. There are two commits, 6f26bb1b and dfe80bc1. Each has one parent and touches only round-k1-rdy-1/preregistration.md and amendments/A1–A9.
• I called the verifier's own control_plane(D, head) and check_t1(D, head) at both commits: no codes. The record directory resolves to verification/programmes/oi-qm/reconstruction/round-k1-rdy-1/; the round is K1-RDY-1, non-sealing.
• Separately, I wrote an independent byte-level scan that does not use the verifier. It finds exactly one v3-round opener (preregistration line 383) and exactly one v3-governed-paths opener (line 389), both closed. It finds no near misses, no CR bytes and no blocks in any amendment.
• The governed paths are just record AM <record dir> and record AM verification/receipts/K1-RDY-1.json. No execution entry is declared. I recomputed the governed-path digest independently: c7ecedf33fdab25da74d9ac29e96c13621445d05caafb30d651e869c01300d1d, which matches. The record directory and the receipt path are both unused at D.
• Checks, all exit 0, full history:
  • v3_verifier.py --self-test: OK.
  • --corpus: 140 vectors, exact and as expected.
  • --receipts <head>: 52 receipts, all hold.
  • legacy_records_check.py <head>: 303 records in 75 closed namespaces, intact.
  • citation_roles_check.py <head>: all hold.
  • artifact_placement_check.py, duplicate_check.py, staleness_check.py (13 matched) and build_migration_manifest.py --check: OK.
  • SC-42 integration check: PASS, 0 added-text hits. Its output and its controls match the checked-in JSON byte for byte.
  • My own scan of the D..head added lines finds no local filesystem path, no session-workspace word, no secret and no e-mail address.

2. Protected text

• head -n 313 … | sha256sum gives a93f3f33…bb63 at both 6f26bb1b and dfe80bc1.
• 6f26bb1b's round directory has tree b93d15f2…, identical to proposed-k1-rdy-1/ at 66bee341. That is byte identity.

3. Faithfulness of dfe80bc1

Every hunk traces to one of three sources:

• main moving: evidence(#1031 K1-RDY): written proofs of the blindness lemma, P1 and L-O(i)/(ii) (WIP / UNVERIFIED) #1119 → 4f016afd, evidence(#1031 K1-RDY): land PROOFS.md and record the #1119 review (OD-1, OD-2) (WIP / UNVERIFIED) #1122 → 4b4b437c;
• research(#1031 K1-RDY): recovery of the accepted preregistration (WIP / UNVERIFIED; not frozen) #1114 c.6653656756 (OD-1…OD-7, 08:44:44Z);
• K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6655647525 (drafting approval, 10:27:34Z).

I compared each quote with the API body; each is verbatim and carries author and time. Specific points:

• Prospective ratification: kept as prospective (§12; A1 "not attributed to the owner for any date before 2026-11-05").
• D-07(a) and O-3: carried as ruled (A1 "Adopted form").
• A5: keeps the quoted condition "Interpretations that depend on a proof remain conditional until those proofs pass review". It dropped the draft's "support this round as proved results once…", which is the narrower choice.
• OD-4: paper-only, matching c.6653410073's option. No path is reserved.
• A4 heads: I re-read them at 10:43Z. All six equal the drafting values, and Introduction paper: reconcile sources and integrate scoped theorem recoveries #918's 44236326 descends from fa32b043.

4. §A.42 item 8 (A7)

I classified every off-main reference in lines 1–313, §12 and A1–A9 myself, and I agree with A7's classes, with one exception (R1).

• Every full commit id in the diff exists and is reachable from main: 19010635, d8bfd07b, 6525ad34 and 4f016afd (all ancestors of 4f016afd), and 4b4b437c.
• 07d94f2c, 02009693 and af1e9be2 are on main. e02544d5 (RR-1), 66bee341 and the PR heads are off main. They are correctly classed as provenance; the rules they would support are stated in full in the text.
• LEMMAS.md, PROOFS.md and both REVIEW.md files are on main at D, entered as §12 says.
• They support the uses as made:
  • Lemma P1, blindness, L-O(i) and L-O(ii) → A5;
  • IND → A3;
  • §1/§4/§5 → A6;
  • CR-3…CR-6 → A8, with CR-5 left open as stated;
  • CYC (156 cycles) → line 191/270's member list;
  • LEDGER → (c1)/(c2).

Two limitations of that support are worth stating in A7:

• evidence(#1031 K1-RDY): written proofs of the blindness lemma, P1 and L-O(i)/(ii) (WIP / UNVERIFIED) #1119's record on main is the coordinating session's condensed posting, which the reviewer confirmed in c.6653843874. It is not the reviewer's verbatim report.
• PROOFS.md's full review happened on an evidence branch (provenance). Its on-main support is the hand re-derivation summary in REVIEW.md §1.

Required before F

• R1 (A7 Main row, and the PR body): misattribution. The substantive change at Main 744/756/758 entered main in 77edc1a1, "Scope fast-bath exclusions to the erasure bound and the timescale form of C2 (Correction planning: scientific overclaims and status corrections (from the 2026-11-01 disposition) #1028 unit 1)", merged via Fast-bath correction: suppression not exclusion; C2-slow vs C2-structural; P-divisibility scope (#1028 unit 1) #1035 on 2026-11-02 at 08:24Z, after the base's acceptance. That change:

  • reads (C2) as (C2-slow);
  • makes (C2-structural) and (C4) suppressed, not excluded;
  • drops "fails (C4)" and "M_t = 0".

  ARROW-1: is the arrow of time derived, or inherited from time-asymmetric initial conditions? (cross-document disagreement) #1112 (d3e13c99) only retitled the corollary and the 744 label phrase ("statement and hypotheses unchanged"). Correct the attribution. Item 8 also says that support on main must state any difference from the cited version, so add one sentence on the difference's effect on the claim. See assessment (b).

• R2 (G6 under OD-4). Base line 150 says "For enumerated rows (including P1 member rows), 'independently' means a second, separately written enumeration". Under paper-only execution no row is enumerated. It is therefore undecided whether a positive P1 member row needs only a second, independent written derivation, or ends in UNDECIDED(G6). The choice changes tokens. Resolve it explicitly before F, by an owner ruling carried in a scoped amendment, as was done for A-1.

• R3 (§12, §A.41). "No enumeration tool and no workflow path is declared, so none can be used in execution" claims more than S7 measures:

  • S7 constrains only which paths change;
  • record AM on the record directory admits any file inside it, a script included;
  • a computation need not be committed at all.

  Either give paper-only (OD-4) a checkpoint, or narrow the sentence to what S7 checks. A possible checkpoint: at E, the reviewer checks that delta(F, E) adds only result.md, gate-table.json and named derivation .md files, none executable, and that the execution log records no computation.

• R4 (A4). The heads placeholder ‹to be re-read …› must be filled in the last control-plane commit, and the independent exact-head review and the F check-run must cover that commit.

Owner decision before F

• R5 (A5's condition). See assessment (a). If the owner rules that the condition is not yet met, tokens that rest on the lemmas need a way to carry it, for example a named hypothesis in Λ or a field in A9. Otherwise a conditional result would be issued unconditionally.

Non-blocking

• A6 / line 252. The "drafting log" has no designated location. Name it, for example: this PR's commits and checkpoint comments, quoted in result.md's first section.
• §12. It could name OD-7 for completeness; OD-7 is administrative.

Assessment (a): do model-authored reviews on main satisfy "pass review"?

• What the rules distinguish. §A.17 makes peer review the verification stage. §A.23 says only external review or replication banks a result, and internal work, however careful, moves consistency only. §A.46 item 1 asks that evidence meet "the verification standard for its kind". Item 7 rules out "an unverified record, an author-posted acceptance, an Issue, or the mere presence of a file on main", but not a recorded review by a non-writer. Item 6 says the lack of a separately named reviewer does not by itself make verified work unverified.
• What the record contains. Both reviews are recorded on main, were done by agents other than the writer, re-derive each step by hand, and check each equivalence in both directions (§A.34). The lemmas' layer is "mathematical derivation on paper, independently reviewed (model-authored)". It is not external peer review, and not kernel-checked.
• What the owner has already said. OD-2's stated rationale, "Without this the lemmas stay conditional under §A.46", which the owner approved, reads as the owner accepting a model-authored review recorded on main as the route out of conditional status.
• My assessment. On the corpus rules the condition can be met, at the internal-verification level and no higher. Three things should be stated:
  1. CR-5 stays open whatever the ruling; A8 already says so.
  2. Nothing here banks correctness (§A.23).
  3. The ruling itself is the owner's to make; A5 should quote it rather than imply it.

Assessment (b): Main §4.6 after the anchor

• Who changed it. The change is Correction planning: scientific overclaims and status corrections (from the 2026-11-01 disposition) #1028's (77edc1a1), not ARROW-1's (R1).
• No gate, token or hypothesis label moves.
  • P3's rule part is fixed by K-12(b).
  • Tokens are computed from the rule part, φ and the measures (§5.2–§5.4).
  • P3's rows are fixed in advance from the image statement (line 235).
  • §4.6 enters a row only through named hypotheses, "(EM) with τ_mix ≪ τ_S" and H-sel, whose labels are unchanged.
• A4's safeguard does not apply by its terms. It covers referenced historical PR heads, not pinned manuscript anchors.
• A precision about the writer's reading. Base line 81 calls Main:62 "motivation only". §4.6 is P3's stated anchor, under (EM), which is open for linear K1 (Main:760, unchanged).
• The anchor is weaker for P3's event. At D the rule excludes only (C2-slow) observers, the timescale form. Structural persistence and (C4) readback are only suppressed. P3's O, configuration identity over three ticks, is a structural-persistence event, so its anchor is weaker at D. The 07d94f2c text's "fails (C4)" and "M_t = 0" were corrected on main, and §A.46 item 7 bars using a corrected statement as true in a premise.
• What I recommend.
  • No new gate amendment is needed.
  • A7's Main row should add one sentence: P3's anchor supports the row only as the (C2-slow)-scoped observer-selection rule at D, under (EM); the 07d94f2c wording's (C2)/(C4) exclusion is not used; no gate or token reads §4.6.
  • If the owner prefers the anchor read at D, that becomes a one-line scoped amendment of line 81's anchor clause.

CI at the head

Run 42012181967 (verify, pull_request, head_sha dfe80bc1) has conclusion success:

• 25 of 26 jobs succeeded, including "Mathlib bridge" (whose Release gate step succeeded), "Lean kernel check" and "Numerical probes". "Numerical probes" is the aggregate job; its step "Require every probe shard" succeeded.
• 1 job was skipped: "A42 exclusion", which is workflow_dispatch-only.

This is a pull-request run of a synthetic merge, so it is not an F attestation (§A.39).

---

Generated by Claude Code

[amaybaum]

Correction to the review above (c.6656052333), relayed for the reviewer. Under R2, G6's clause "For enumerated rows (including P1 member rows)…" is at base line 151, not line 150. Nothing else in the review changes.

---

Generated by Claude Code

[amaybaum]

Checkpoint 52ae69fb9731901dac8af228fbe1f930adc23fd1 (single-parent on dfe80bc1, control-plane files only): applies R1–R5 and the minor items of review c.6656052333 under the owner's rulings #1031 c.6656166518 (new A10, A11; A4 heads filled at 2026-11-05T10:57:50Z; first 313 lines sha256 a93f3f33…bb63 unchanged; T1 predicates no codes; SC-42 PASS). Awaiting an independent delta review; no F, dispatch, enumeration or execution.

---

Generated by Claude Code

[amaybaum]

Delta review dfe80bc1 → 52ae69fb9731901dac8af228fbe1f930adc23fd1 (pre-F draft)

This review was read-only and model-authored, by the same reviewer as c.6656052333 and not the writer. I used a fresh full clone with no alternates and deleted it afterwards. I ran no enumeration or model computation.

Verdict: ACCEPT WITH CHANGES.

• R1, R2, R3, R4 and the minor items are applied as ruled in K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6656166518.
• R5 is applied for LEMMAS.md. Two precision fixes are needed in A5 (item F1 below).
• The ruling's scope over the PROOFS.md propositions needs to be settled before F (item F2 below and §5).

1. Lifecycle and checks

• Commit shape. 52ae69fb has a single parent, dfe80bc1. It touches only control-plane files:

  • preregistration.md;
  • amendments A4, A5, A6 and A7;
  • the new amendments A10 and A11.

• Protected text. head -n 313 | sha256sum still gives a93f3f33…bb63.

• Verifier, called directly. control_plane(D, head) and check_t1(D, head) report no codes. The record directory, K1-RDY-1 and non-sealing are unchanged. The governed-path digest is still c7ecedf3…0d1d.

• Independent byte scan over the 12 control-plane files:

  • exactly one v3-round block (line 388) and one v3-governed-paths block (line 394);
  • no near misses, no CR bytes, and no blocks in any amendment.

• Tool checks, full history, all exit 0:

  Check	Result
  --self-test	OK
  --corpus	140, exact and as expected
  --receipts <head>	52 hold
  legacy_records_check.py	303 intact
  citation_roles_check.py	all hold
  artifact placement	OK
  duplicate	OK
  staleness	OK, 13 matched
  build_migration_manifest --check	OK
  SC-42 integration	PASS, 0 added-text hits; result and controls match the JSON byte for byte

• Added-text scan. My own scan of the added lines finds no local filesystem path, no session-workspace word, no secret and no e-mail address.

• Identifiers. Every new full commit id is reachable from main: 77edc1a1, 01a096a9 (which merges 77edc1a1), d3e13c99, 6ff20b91 and 022ae8da. 06d6ab31 and 7c768f05 are off main and are correctly classed as provenance.

• Owner quotes. The R2, R3, R4 (partial, so marked) and R5 quotes match c.6656166518, with author and time.

2. R1–R5 and the minor items

R1: applied.

• A7's Main row now attributes the change to 77edc1a1 (Correction planning: scientific overclaims and status corrections (from the 2026-11-01 disposition) #1028 unit 1), merged by Fast-bath correction: suppression not exclusion; C2-slow vs C2-structural; P-divisibility scope (#1028 unit 1) #1035 in 01a096a9 on 2026-11-02.
• It records that ARROW-1: is the arrow of time derived, or inherited from time-asymmetric initial conditions? (cross-document disagreement) #1112 (d3e13c99) only retitled the corollary.
• It states that P3's anchor relies only on the (C2-slow)-scoped rule at D under (EM), and that no gate or token reads §4.6.
• No base line, gate, token, control or hypothesis label changed. P3's Λ still names "(EM) with τ_mix ≪ τ_S".

One non-blocking point: A7's header says "Governs: no base line's content", but the R1 sentence in effect reads line 81's §4.6 anchor at D rather than at 07d94f2c. The owner directed that reading, so A7's Governs line could say "line 81's anchor clause: reading only".

R2: applied (A10).

• A10 governs base line 151 (continued on 152) without editing it: the 313-line hash is unchanged, and the rest of G6, lines 150–153, stands.
• Its three conditions can be checked at E:
  1. not a restatement, and none of the first derivation's intermediate results taken as given;
  2. the same claim under the same assumptions, namely the frozen rule and Λ;
  3. the same token as the adversarial note of lines 150–151, with G2(b) recomputed.
• Otherwise the row ends in UNDECIDED(G6), and a paper derivation is never presented as computational or kernel-checked.

Non-blocking: "genuinely independent" could be made more checkable by having each derivation file record its author or session, and whether the second writer had access to the first.

R3: applied (A11; §12 lines 383–385).

• A11 adds a row to §9 (lines 248–270) and is described as a reviewer checkpoint under §A.41.
• It states explicitly that it "is not a predicate of tools/v3_verifier.py" and that record AM admits any file.
• It states that the log's no-computation statement "is a declaration by the executor; by itself it is not independent proof".
• Its concrete clauses are: the path classes, mode 100644, no script, program, notebook or workflow, and no execution path. A violation blocks E pending an owner decision.
• §12 is narrowed to what S7 checks.

Non-blocking: §12's "it is the checkpoint of A11" could read "it is checked by the reviewer checkpoint of A11", so that it echoes A11's own limitation.

R4: applied (A4).

• The heads are filled with full ids, read at 10:57:50Z.
• I re-read them independently at 2026-11-05T11:04:35Z, and all six are identical:
  • 1009 d4a80fc8…
  • 1023 af1e9be2…
  • 1011 947761dd…
  • 1003 0efd67a6…
  • 918 44236326…
  • 1025 e02544d5…
• A4 states that the exact-head review and the F check-run cover this commit, and that any later control-plane commit re-reads the heads.
• For information: main has moved past D, to f5efa7a0. #918 landing: the Introduction (five files + six ledger rows), with the SC-42 REST allowance #1124 landed Introduction paper: reconcile sources and integrate scoped theorem recoveries #918's Introduction files there. Introduction paper: reconcile sources and integrate scoped theorem recoveries #918's head is unchanged and no anchor of the round depends on it, so this is not an A4 mismatch.

R5: partly applied (A5 lines 17–41; A7).

• What is correct:
  • the reviewed statements and commit 19010635;
  • the change and check chain d8bfd07b → 6525ad34;
  • LEMMAS.md at D differs from 6525ad34 only in its five status lines (checked);
  • the condensed-versus-verbatim distinction for 6651717259;
  • the ruling quoted, its scope limits, and CR-5 kept open;
  • A7's row on the PROOFS.md review provenance. I checked it: the PROOFS §1–§6 text is unchanged from 06d6ab31 to 7c768f05, the 7c768f05 blob 3801aba6 is the one landed in 022ae8da, and the version at D differs from it only in its two status lines.
• F1 (fix before F): two precision points, under the ruling's own proviso that "provenance and exact reviewed statements are established".
  • Lines 21–22 attribute to the review "no change to a statement or proof step, apart from one clause added to Lemma P1(c)". The review says "No statement or proof step changes". The P1(c) clause is the review's optional addition, and REVIEW.md notes it separately. The text should be split accordingly.
  • Line 29 says comment 6653737476 "is the reviewer's report word for word". Per k1-rdy-proofs-20261105/REVIEW.md, that holds only between its BEGIN and END markers, apart from the marked omission. The comment's header is the coordinating session's. Its dated statement after END "condenses a statement the same reviewer wrote at posting time". Lines 24–25 rest the confirmation of 6525ad34 on that statement, so they should say it is condensed.
• F2 (owner clarification before F): the ruling's scope over PROOFS.md. See §5.

Minor items: applied.

• A6 names the drafting log: the D..F commit history of the branch plus the checkpoint comments on K1-RDY-1: native V3 round, halted after P0 (paper-only; S12 withdrawal and halted receipt) #1126, quoted in result.md's first section.
• §12 cites OD-7.
• §12 points to where each ruling is quoted.

3. A11

A11 is described accurately as a new reviewer checkpoint, not a verifier guarantee. It does not overclaim from the log declaration.

4. A10

A10 governs base line 151 by reference, not by edit. Its independence conditions can be checked at E against the named derivation files, subject to the authorship suggestion above.

5. Assessment: does R5 cover only LEMMAS.md, or also the PROOFS.md propositions?

The ruling's text supports the broader reading.

• It says "reviews", plural, "of the lemmas on main". There are two review records on main: k1-rdy-proofs-20261105/REVIEW.md and k1-rdy-pref-proofs-20261105/REVIEW.md. The evidence(#1031 K1-RDY): written proofs of the blindness lemma, P1 and L-O(i)/(ii) (WIP / UNVERIFIED) #1119 review is one review plus its delta check.
• The same ruling orders the writer to "record that the full PROOFS.md review was performed on an evidence branch". That is a provenance proviso, which only matters if the PROOFS.md review is among the reviews the ruling relies on.
• The record names these items as lemmas: evidence(#1031 K1-RDY): land PROOFS.md and record the #1119 review (OD-1, OD-2) (WIP / UNVERIFIED) #1122's merge on main is titled "K1-RDY preference-lemma proofs and review evidence", and K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6654995013 says the same.

What the narrow reading (LEMMAS.md only) would change.

• The owner's general condition, "Interpretations that depend on a proof remain conditional until those proofs pass review", would leave these conditional: IND, CYC, EN, LEDGER, E (§1), P3's image (§3) and CR-1 to CR-4 and CR-6.
• Those carry these parts of the round:
  • S0 control values (A6: E, CC-ABS's r(w), CC-LEDGER's shells);
  • P0's B-eq row (IND);
  • the P1 member list and the "No post-selection" checkpoint (CYC);
  • the G3 bounds (LEDGER);
  • P3's fixed rows (its image);
  • A8's scope (CR).
• The base has no way to carry that condition. Controls would be "passed conditionally", which has no token, and every token would inherit an unstated condition. So the narrow reading would need either a further amendment that carries the condition, for example a named hypothesis in Λ and on S0 rows, or a separate owner ruling.

What the broad reading would change.

• It needs one paragraph, in A5 or beside A7, that establishes the PROOFS.md reviewed statements exactly, as the ruling's proviso requires:
  • §1–§6 reviewed at 06d6ab31 (research(#1031 K1-RDY): recovery of the accepted preregistration (WIP / UNVERIFIED; not frozen) #1114 c.6653236579), with text unchanged in 7c768f05;
  • §7 (LEDGER) reviewed at 7c768f05 (c.6653410073);
  • the reviewed blob 3801aba6 is the one landed, and the version at D differs only in its two status lines;
  • the on-main support is REVIEW.md §1.
• It would then state that the condition is met at the internal-verification level for E, CYC, P3's image, IND, EN, CR-1 to CR-4, CR-6 and LEDGER, with CR-5 still open.
• Nothing else in the round changes.

My recommendation. The owner confirms the broad reading, and the writer adds that paragraph before F. A5's present text, "met for these statements", where "these" means only the four LEMMAS.md results, leaves the propositions the round's controls depend on silently outside the ruling.

6. CI at the head

Run 42015176647 (verify, pull_request, head_sha 52ae69fb) completed with conclusion success:

• 25 jobs succeeded, including "Mathlib bridge", whose "Release gate" step succeeded, "Lean kernel check", and the "Numerical probes" aggregate.
• 1 job was skipped: "A42 exclusion", which runs only on workflow_dispatch.

This run tests a synthetic merge with current main (f5efa7a0), so it is not an exact-head F attestation (§A.39).

Before F: apply F1, get the owner's answer on F2 and add the corresponding text, then run a delta check of that commit. If that commit is the last control-plane commit, re-read A4's heads in it.

---

Generated by Claude Code

[amaybaum]

Checkpoint 9fa37d0a1aa69e7cebcced971b6fb627e8724d96 (single-parent on 52ae69fb, control-plane files only): applies F1/F2 of delta review c.6656432806 under the owner's ruling #1031 c.6656645169 (A5 wording fixes; PROOFS.md review provenance and exact reviewed statements; internal-verification scope only, CR-5 open), the three optional one-liners, and A4 heads re-read at 2026-11-05T11:20:15Z (unchanged). First 313 lines sha256 a93f3f33…bb63 unchanged; T1 predicates no codes; SC-42 PASS. Awaiting an independent delta review; no F, dispatch, enumeration or execution.

---

Generated by Claude Code

[amaybaum]

Delta review 52ae69fb → 9fa37d0a1aa69e7cebcced971b6fb627e8724d96 (pre-F draft)

This review is read-only and model-authored. I am the reviewer of c.6656052333 and c.6656432806, not the writer. I worked in a fresh full clone without alternates and deleted it afterwards. No enumeration or model computation was run.

Verdict: ACCEPT. F1 and F2 are applied correctly, under #1031 c.6656645169. No statement, hypothesis, gate, token or conclusion changed. In my view this head is ready to present to the owner as the candidate F. The steps still needed before F are listed at the end.

1. F1: A5 lines 19–35

Both wording fixes match the review records.

• Lines 21–24. The review says "no statement or proof step changes". The P1(c) clause "hence φ^k(supp μ) = supp μ for all k ∈ ℤ" was the review's optional addition, adopted by the writer. The delta check verified the clause and its proof clause: k1-rdy-proofs-20261105/REVIEW.md, delta report "P1 (iii): applied … The added proof clause … is correct and sufficient".
• Lines 26–28 and 32–35. Only the text between BEGIN and END in comment 6653737476 is verbatim. The header belongs to the coordinating session. The dated statement after END is a condensation, and the confirmation of 6525ad34 rests on it. All of this matches REVIEW.md's own account ("condenses a statement the same reviewer wrote at posting time").

2. F2: A5 lines 37–82, checked independently

Provenance

• 06d6ab31f7f3… and 7c768f0590d9… are the full ids. Both are on evidence/k1-rdy-pref-proofs-20261105, off main, and are correctly marked as provenance.
• Between them, PROOFS.md gains exactly one row in the §0 table (line 24, the (c1)/(c2) row) and §7. §1–§6 are unchanged.
• The blob at 7c768f05 is 3801aba6c972…. It is identical to the blob landed in 022ae8da.
• At D, the file differs from it only in its two status lines (5 changed lines, two hunks).
• 6ff20b91 is on main, and its REVIEW.md verdict is ACCEPT.

Covered statements. Each listed item appears in the review records: #1114 c.6653236579 for §1–§6, c.6653410073 for §7, and REVIEW.md §1. The list adds nothing beyond them.

• §1: E and W_eq (base line 198 checked).
• §2: CYC — 156 cycles, period counts, φ^{10}=id, and the short cycles.
• §3: Im_O — index 8, π_0(O)=1/8, D(μ_O)=3 log 2.
• §4: IND and its three uses.
• §5: EN (a)–(c).
• §6: CR-1–CR-4, CR-6, and the reading of "every weighting supported on C".
• §7: LEDGER, and its consequence that VOID[row] arises only from a false readiness assertion.
• CR-5 is recorded as "checked as stated": the converse of CR-3 is not established, and the converse stays open. This matches c.6653236579 ("The rule is sufficient, not an equivalence") and REVIEW.md §1.

Ruling and scope

• I compared both quotes line by line with the API bodies. The F2 quote (c.6656645169, 11:18:22Z) and the R5 quote (c.6656166518) are verbatim, with author and time.
• Lines 77–82 limit the effect to "the statements listed under 'Statements reviewed' … and for no others".
• The disclaimers are all present: not external peer review; not external or independent human validation; not Lean verification; not proof that the physical assumptions follow from OI. CR-5 stays open.

3. Preservation and lifecycle

• 9fa37d0a has a single parent, 52ae69fb. It changes only control-plane files: preregistration.md, A4, A5, A7 and A10.
• head -n 313 | sha256sum gives a93f3f33…bb63.
• My independent byte scan finds exactly one v3-round block (line 388) and one v3-governed-paths block (line 394). It finds no near misses, no CR bytes and no blocks in any amendment.
• The governed-path digest is unchanged: c7ecedf3…0d1d.
• No mathematical statement, hypothesis, gate, token, control or conclusion is touched. The changes are confined to provenance, the scope text, the A4 timestamp and three wording edits.

4. Optional edits

• A7 line 3 is meaning-neutral. It discloses the Main-row reading of line 81's anchor at D that the owner directed in R1. No gate reads it.
• Preregistration line 385 is meaning-neutral. It now reads "checked by the reviewer checkpoint of A11", which matches A11's own scope.
• A10 line 16 is not strictly meaning-neutral. It adds a recording obligation: each derivation records its author or session, and whether the second writer had access to the first. It changes no gate, token or pass condition, and it is always satisfiable. It makes the owner's "genuinely independent" condition checkable at E. It goes beyond the literal F1/F2 authorization, though, so it is flagged here for the owner's awareness. It is not a defect.

5. A4 heads

I re-read the heads independently with git ls-remote at 2026-11-05T11:24:39Z, after the writer's 11:20:15Z read. All six are identical to the A4 table:

PR	Head
1009	d4a80fc8…
1023	af1e9be2…
1011	947761dd…
1003	0efd67a6…
918	44236326…
1025	e02544d5…

6. Checks on full history (all exit 0)

• Verifier internals: control_plane and check_t1 on (D, head) return no codes.
• v3_verifier.py:
  • --self-test: OK.
  • --corpus: 140, exact and as expected.
  • --receipts <head>: 52 receipts hold.
• Repository checks:
  • legacy_records_check.py: 303 intact.
  • citation_roles_check.py: all hold.
  • Artifact placement, duplicate and staleness checks: OK (staleness matched 13).
  • build_migration_manifest --check: OK.
• SC-42: PASS with 0 added-text hits. The result and the controls match the JSON byte for byte.
• Added-text scan: no local filesystem path, no session-workspace word, no secret and no e-mail address.
• main has moved to 5e35eb88 (research(#1033 BSEL-2): unit B4′, gauge equivalence vs complete-record operational equivalence #1123 and #918 landing: the Introduction (five files + six ledger rows), with the SC-42 REST allowance #1124). Neither touches tools/, .github/, AGENTS.md or this round's directory, so the round is unaffected. The only reconciliation point comes after E.

7. CI at the head

Run 42018230253 (verify, pull_request, head_sha 9fa37d0a) concluded success.

• 25 jobs succeeded, including "Mathlib bridge" (its "Release gate" step succeeded), "Lean kernel check" and the "Numerical probes" aggregate.
• 1 job was skipped: "A42 exclusion", which runs only on workflow_dispatch.
• This run tests a synthetic merge with current main, so it is not the F check-run attestation.

Before F (each step needs the owner's separate approval)

1. Hold the branch at exactly 9fa37d0a from the start of the candidate-F hold until the designation or an explicit release (§A.43 item 3). Any further control-plane commit would need a re-read of the A4 heads and a new delta review.
2. Run an exact-head check-run. This is a workflow_dispatch of verify on the branch held at 9fa37d0a. The run whose head_sha is 9fa37d0a, with every job green, is F's attestation (§A.39). The dispatch is a launch and needs owner authorization (§A.43 item 4).
3. The owner designates F = 9fa37d0a…, citing that run (K-11).

Execution after that still needs the separate #1031 gate-3 authorization.

---

Generated by Claude Code

[amaybaum]

Checkpoint d9b3de012e393262adb20f91ad49627bad48de43 (single parent 9fa37d0a, control-plane files only; hold released for this one commit per #1031 c.6657384912): A10 third recording item with the permitted-inputs/intermediate-results distinction (Option (b), quoted with A10 KEEP), A4 heads re-read at 2026-11-05T11:58:40Z (unchanged), A7 trace rows. First 313 lines sha256 a93f3f33…bb63 unchanged; T1 predicates no codes; SC-42 PASS. The 9fa37d0a run is evidence for that historical head only. Awaiting an independent delta review; no F, dispatch, enumeration or execution.

---

Generated by Claude Code

[amaybaum]

Delta review 9fa37d0a → d9b3de012e393262adb20f91ad49627bad48de43 (pre-F draft)

This is a read-only, model-authored review by the same reviewer as the previous three reviews, not the writer. I used a fresh full clone with no alternates and deleted it afterwards. I ran no enumeration or model computation.

Verdict: ACCEPT. d9b3de01 is ready to present to the owner as the revised candidate F. What remains before F is the exact-head verify.yml dispatch on this SHA and the owner's designation.

1. A10 (lines 15–27 and 34–48)

• Recording items. Lines 16–20 list all three:
  • the author or producing session;
  • whether the second writer had access to the first derivation;
  • any relevant dependency or shared intermediate result, "recording 'none' explicitly where there is none".
• Permitted inputs versus dependencies (lines 22–23):
  • The shared permitted inputs are named: the frozen rule, the declared hypotheses (the row's Λ) and the row specification. They "are not dependencies under item 3".
  • "an intermediate result taken from the first derivation is" a dependency.
• Access alone (lines 23–24): "Access to the first derivation does not by itself establish dependence."
• Reliance (lines 24–25): relying on an intermediate result of the first "does not meet the first condition above, so the row ends in UNDECIDED(G6)". This matches the unchanged bullet at line 10 ("takes none of its intermediate results as given") and the UNDECIDED(G6) outcome at line 15.
• Procedural only (lines 25–26): the records "introduce no gate or token, and recording them does not by itself establish independence". Lines 7–15 are unchanged, and the paper-derivation clause at lines 26–27 is kept.
• Owner quotes. I compared each quote with the API body; all are verbatim, with author and time.
  • K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6657308064 (amaybaum, 11:53:35Z) is quoted in part: its "A10 — KEEP" paragraph.
  • K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6657384912 (amaybaum, 11:57:45Z) is quoted in full: all eight quoted lines.
  • The text implements c.6657384912 item by item.

2. Preservation and lifecycle

• Commit shape. d9b3de01 has a single parent, 9fa37d0a. It changes only A4, A7 and A10, all control-plane files. preregistration.md is untouched.
• Protected text. head -n 313 | sha256sum gives a93f3f33…bb63.
• Declaration blocks. My independent byte scan finds exactly one v3-round block (line 388) and one v3-governed-paths block (line 394). It finds no near miss, no CR and no block in any amendment.
• Governed-path digest. Unchanged: c7ecedf3…0d1d.
• Scientific content. No hypothesis, gate, token, control or conclusion changed. The only changes are A10's procedural records, A4's read timestamp and three A7 trace rows.

3. A4 and A7

• A4. I re-read all six heads independently with git ls-remote at 2026-11-05T12:03:27Z, after the writer's read at 11:58:40Z. All match the A4 table:

  PR	Head
  1009	d4a80fc8…
  1023	af1e9be2…
  1011	947761dd…
  1003	0efd67a6…
  918	44236326…
  1025	e02544d5…

  The branch ref is at d9b3de01.

• A7, lines 46–48. All three classifications are correct:

  Reference	Classified as	Why it is correct
  c.6657384912	"support on main once A10 lands"	quoted in A10 with author and time
  c.6657308064	support once A10 lands, for its A10 paragraph	that paragraph is quoted
  c.6657308064	provenance, for the dispatch approval at 9fa37d0a	that is a historical head
  K1-RDY-1: native V3 round, halted after P0 (paper-only; S12 withdrawal and halted receipt) #1126 c.6656892864	provenance	the owner's decisions on it are quoted

4. Full-history checks (all exit 0)

• Control plane. control_plane and check_t1 on (D, head) report no codes. The round is K1-RDY-1, non-sealing.
• Verifier:
  • --self-test: OK.
  • --corpus: 140 vectors, exact and as expected.
  • --receipts <head>: 52 receipts, all hold.
• Repository checks:
  • legacy_records_check.py: 303 records intact.
  • citation_roles_check.py: all hold.
  • artifact placement and duplicate checks: OK.
  • staleness check: OK, 13 matched.
  • build_migration_manifest --check: OK.
• SC-42. PASS, with 0 added-text hits. Its result and controls match the checked-in JSON byte for byte.
• Added-text scan. No local filesystem path, no session-workspace word, no secret and no e-mail address.

5. CI

• PR run at this head. Run 42023394557 (pull_request, head_sha d9b3de01) concluded success.
  • 25 jobs succeeded, including "Mathlib bridge" with a successful "Release gate" step, "Lean kernel check" and the "Numerical probes" aggregate.
  • "A42 exclusion" was skipped; it runs only on workflow_dispatch.
  • This run tests a synthetic merge, so it is not an F attestation.
• Earlier dispatch. Run 42022636754 (workflow_dispatch at 9fa37d0a) was still in progress when I checked. Per c.6657384912, it is evidence for that historical head only.

Before F (each step needs separate owner approval)

1. Hold the branch at exactly d9b3de01. The owner's c.6657384912 authorized this commit while the 9fa37d0a hold of c.6657308064 was in force; I read that as releasing the hold. If the owner wants an explicit release recorded, that is a one-line note (§A.43 item 3).
2. Dispatch verify.yml once on the held branch. The run whose head_sha is d9b3de01, with every job green, becomes F's check-run attestation (§A.39). The dispatch itself needs the owner's approval (§A.43 item 4).
3. Owner designates F = d9b3de012e393262adb20f91ad49627bad48de43, citing that run (K-11).

Execution after that still needs the separate #1031 gate-3 authorization.

---

Generated by Claude Code

[amaybaum]

Exact-head verification result for the candidate F d9b3de012e393262adb20f91ad49627bad48de43. This was run under the owner's approval at #1031 c.6657843984.

• Run. verify.yml run 42026370642, event workflow_dispatch, ref round/k1-rdy-1-20261105.
  • Every job reports head_sha = d9b3de012e393262adb20f91ad49627bad48de43. As a dispatch run, it tested the exact commit, not a synthetic merge.
  • Run conclusion: success.
• Jobs. All 40 of 40 succeeded, read from structured job and step conclusions:
  • the Lean kernel check;
  • the Mathlib bridge, including its Build and Release gate steps. The Release gate covers the v3-receipts, legacy-records and control-plane steps.
  • every numerical-probe shard and the aggregate job;
  • all 15 dispatch-only A42 exclusion shards: cubes:0–2, dfs:0–5, n18:0–2, paths, r5 and sathr.
• Skips. No job was skipped. The only skipped step is the Mathlib bridge's "KTRANS-SEP-2 marker complete reports with an axiom outside the three". It is a failure marker that runs only when an unexpected axiom is found, so skipping it is the passing case.
• Hold. git ls-remote after the run still shows the branch at d9b3de01. Nothing was modified.
• Preservation. From the delta review c.6657687217:
  • the first 313 lines still hash to a93f3f33…bb63;
  • there is one v3-round block and one v3-governed-paths block;
  • the governed-path digest c7ecedf3…0d1d is unchanged;
  • 303 legacy records are intact and 52 receipts hold.
• Earlier dispatch. Run 42022636754 at 9fa37d0a (success) is historical evidence for that head only.

d9b3de01 is presented to the owner for a separate F designation decision. F is not designated by this record.

---

Generated by Claude Code

[amaybaum]

Execution checkpoint 291925dacccc949bae57d465b932aac3fd8d1a25 (linear from F d9b3de01 via 3efb9233; record paths only; WIP / UNVERIFIED), under the limited gate-3 authorization #1031 c.6658964414: S0 complete — all 11 controls pass on paper (one written derivation each under derivations/; gate-table.json controls; result.md §1–§2), so VOID[stage] is not issued; one reading marked (CC-IMP subtype from G2(a)). A4 heads at start (MCP pull-request read, 13:20:43Z–13:21:46Z) all match. Protected hash a93f3f33…bb63 unchanged; verifier control_plane/check_t1/check_execution no codes; SC-42 PASS. No computation, enumeration, dispatch or E. Next: P0.

---

Generated by Claude Code

[amaybaum]

Execution checkpoint da1d939e52e71110008cd6534a672897c64c3339 (linear from F via 3efb9233, 291925da; record paths only, 16 added files, all mode 100644; WIP / UNVERIFIED), under #1031 c.6658964414: P0 evaluated on paper — B-pl, B-eq and B-ct each ABSENT (tokens in gate-table.json; derivations derivations/P0-*-d1.md); no positive P0 row, so W2/W3 not engaged. Stopped at the P1 boundary: no member row, class token, P3 row or E; the S12 halt awaits the owner's go-ahead. A4 heads re-read at the boundary (MCP, 13:29:54Z–13:30:05Z) all match. Protected hash unchanged; verifier control_plane/check_t1/check_execution no codes; A9 key-set OK; SC-42 PASS. No computation, enumeration or dispatch.

---

Generated by Claude Code

[amaybaum]

Independent review of the K1-RDY-1 execution at da1d939e: ACCEPT WITH CHANGES (all non-material)

Commissioned under #1031 c.6659415927 (part A). The reviewer is independent, model-authored and read-only, and did not write any of the work under review. The report is reproduced below.

---

K1-RDY-1: independent read-only review of the paper-only execution at da1d939e

Objects. F = d9b3de012e393262adb20f91ad49627bad48de43; execution head da1d939e52e71110008cd6534a672897c64c3339. Commits 3efb9233 → 291925da → da1d939e.
Authority. Execution: #1031 c.6658964414. This review: #1031 c.6659415927.
Method. git show/diff/ls-tree/log/merge-base, sha256sum, jq (key sets and sorting only), grep and GitHub MCP issue reads. No script, CAS, enumeration or model computation was run. Every value below was re-derived by hand from the frozen rules (base lines 1–313, §12, A1–A11) and from PROOFS.md/LEMMAS.md at D = 4b4b437c.

Verdict: ACCEPT WITH CHANGES (all changes non-material)

• All 11 S0 controls pass on an independent derivation. All three P0 tokens are correct.
• No material scientific disagreement and no defect in S0 or P0.
• Nothing requires a change to frozen rules.
• The changes in §7 are wording and record-accuracy fixes. They change no outcome, value or token.

1. Setup re-derived (mod 2)

• φ(x,y) = (y, Ay+x); backward, y_i(t−1) = x_i and x_i(t−1) = x_{i−1} + x_{i+1} + y_i.
• Readout: V_{t0} = y_1 and V_{t0+1} = y_0 + y_2 + x_1.
• Removing Σ = {x_0, y_0} gives E = {x_1, y_1, y_2} and W_eq = {x_1, y_1}, in agreement with PROOFS.md §1.
• π_0 facts used: φ-invariance; disjoint cell sets independent and uniform; π_0(E=0 | W=w) = 2^{−|E∖W|} if w is 0 on W∩E, else 0. Each holds by direct counting on the product space.

2. S0: all 11 controls

Control	Derivation	Pass?
CC-ABS	W={x_3,y_3} is disjoint from E; each w has mass 1/4 and r(w) = 2^{−3} = 1/8. No sharp w: ABSENT.	yes
CC-EXACT	(19/20)·1 + (1/20)(1/8) = 152/160 + 1/160 = 153/160 < 1. Not sharp; weak value 153/160.	yes
CC-SCOPE	Under δ_0 the state is 0 at every tick. B-pl: sharp, D = 10 log 2 ≥ 3 log 2, single carrier value, NB fails: BLIND. B-eq: only w=00 has mass and is sharp; log 2 ≤ 10 log 2: BLIND. B-ct: φ^{−1}(0)=0, w=00 has mass 1, cross-time so no void: BLIND.	yes
CC-LEDGER	Translation by (𝟙,0) shifts ℰ by 1 (since A𝟙=0), so each shell has 512 states and D = log 2; (c1) needs 3 log 2 ≤ log 2, false: VOID[row].	yes
CC-IMP	μ_Q uniform on 2^7 states, D = 3 log 2. The rule part names E, so (a) flags; T1 and T2 fire. Subtype: §4(i).	yes
CC-IMP′	(a) the conjunct y_2(t_0)=0 names y_2 ∈ E∖W_eq, flagged; (b) the record/copy clause flags the fan-out ancilla copy.	yes
CC-IMP″	At N=6, O = {x_0=x_1=x_2=x_5=y_0=y_1=0}, fixing x_2(t_0)=0. (a) passes; T1 and T2 fire. P(E=0|O) = 1/2 against π_0(E=0) = 1/4.	yes
CC-PASS	O₁ = {x_1+y_1=0, x_0+x_2=0} equals the stated image, so T2 does not fire; T1 does not fire. B-pl content x_1+y_1; B-eq no content.	yes
CC-T2	Computed image {x_1+y_1=0, x_0+x_2=0} differs from stated {x_1+y_1=0} (witness x_0=1). T2 fires; T1 does not.	yes
POS-X	6 independent conditions, |B_w|=16, mass 1/64. W-certified; w=0 sharp; NB holds via V_{t0+1}=y_0(t_0); Λ = {H-μ, finite-S, H-ret, H-2s, H-cur}; CONDITIONAL[Λ].	yes
CC-BLIND	8 conditions, |B_w|=4, mass 1/256. w=0 sharp; carrier fixed by w, NB fails: BLIND.	yes

No VOID[stage]. Every control derivation file at da1d939e agrees and decides its pass condition without relying on HC-1 or HC-A to HC-D (A6).

3. P0 rows (μ = π_0, empty rule part, image S)

G1 passes; G2: (a), T1, T2 do not flag, H-sel content none; G3: D(π_0) = 0.

• B-pl. π_0(E=0) = 1/8; NB false (readout uniform for every carrier value); not W-certified. ABSENT[π_0; B-pl; H-cur (declared)].
• B-eq. Each w mass 1/4; r(00) = 1/2, other three 0; not W-certified. ABSENT[π_0; B-eq; H-cur (declared), H-2s (declared)].
• B-ct. B_w = {x_0+x_2+y_1 = w_{x1}, x_1 = w_{y1}}, mass 1/4 each; r(00) = r(10) = 1/4, r(01) = r(11) = 0; not W-certified. ABSENT[π_0; B-ct; H-cur (declared), H-2s (declared), H-ret].

The weak maps in gate-table.json match in A2's cell order.

4. Interpretation questions

• (i) CC-IMP's subtype. §5.4 defines only single-subtype row tokens (IMPORTED for a G2(a) flag, DYN for G2(b)); A2's set field is for the P1 class token only; G2 lists (a) first; the pass condition was written for a rule part naming E. The execution issued IMPORTED and recorded the (b) firings: the reading the frozen text supports. The side remark that "carrying both subtypes would also attain it" is unsupported (a two-subtype row token is undefined); non-material (C1).
• (ii) G5–G7 not-reached after G4. Line 120, line 183 ("a row takes the token of the first gate that issues one"), G6's positive-row scope and A9's "not-reached" value make this the required reading; evaluating them could not change the token.
• (iii) H-law/H-sel in rows stopping at G4. Λ is written at G5; the ABSENT token carries [μ; branch; access], not Λ; line 56 independently adds H-law to a non-W-certified row. The hypotheses field holds the declaration part plus access labels, without H-law, described as "not evaluated". No token depends on it (non-material), but the description is inaccurate: w_certified=false brings in H-law and hsel_content=[] excludes H-sel (C2).

5. Tokens against frozen precedence

No VOID[stage] (11/11 pass); no Z1-VOID for P0; G1–G3 issue no token; at G4 no sharp w in any branch, so ABSENT (VOID[row] and BLIND need a sharp w); token == gates.G4 for all rows; no aggregate (line 184); class_tokens is {}; no SOURCED token and nothing above AVAILABLE[H-sel].

6. A11, provenance and scope

• Diff F..da1d939e: 16 files, all added, all under the record directory, all mode 100644: result.md, gate-table.json and 14 derivations/*.md, each named in result.md §7. No code, script, workflow or execution path.
• Protected text: a93f3f33f0c5a6e8b2467312d184fa2d2e915296490948938c4dd7832c64bb63 at both F and da1d939e; preregistration.md and amendments/ blobs identical.
• Commits: single-parent, linear from F; first execution commit 13:22:28Z after the authorization at 13:16:45Z; result.md §1.1 quotes c.6658964414 accurately.
• A10 records: present in all 14 derivation files.
• Gate table (A9): key sets match exactly; schema "k1rdy-gate-table", version 1; hypotheses code-point sorted.
• Computation: nothing indicates enumeration or model computation; every value is reachable by short linear algebra over ℤ/2. As A11 states, the executor's declaration is not independent proof. The disclosed get_commits call is the read-only pull_request_read method; the A4 comparison uses get and head.sha; it affects no result.
• P1/P3: stated as not evaluated; no member row, no class token. The halt is the owner's execution-scope decision (§12 line 329), not a frozen-rule change. ABSENT tokens are scoped to (π_0, branch, access, instance); no impossibility claim, no extrapolation, no E.

7. Requested changes (non-material; none touches frozen text or any token)

• C1. result.md lines 119–121 and derivations/S0-CC-IMP.md lines 39–40: replace the "both subtypes" counterfactual with the grounds (single-subtype row tokens only; G2 decides (a) first; pass condition written for a rule part naming E), or delete it.
• C2. result.md lines 145–147 (and the G5 paragraphs at P0-B-pl-d1.md:65, P0-B-eq-d1.md:61, P0-B-ct-d1.md:60): state that for rows stopping at G4 the hypotheses field is the declaration and access part, not Λ; Λ would add H-law and not H-sel. Alternative: add "H-law" to the three lists.
• C3. result.md lines 157–158 rest the halt rationale on K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6658841280, which is off main (§A.42 item 8). Quote its operative sentences or narrow the line to the owner decision quoted in §1.1.
• C4 (optional). derivations/S0-POS-X.md line 83: "are not run (§9, controls in isolation); Λ is the fixture's stated set."

C1–C3 can be confined to result.md, which W updates. Editing a derivation file or gate-table.json would need explicit owner allowance under step B. Each change is to be independently rechecked (c.6659415927).

8. Explicit flags

• Material scientific disagreement: none.
• Defect in S0 or P0: none. 153/160, 1/256, 1/8, 1/2, 1/4, 1/64 and 3 log 2 confirmed.
• Change to frozen rules required: none.

---

Next (c.6659415927, part B). The conditions for the S12 halt are met: ACCEPT WITH CHANGES, all non-material, no frozen-rule change. The sole branch writer writer-k1rdy1-exec is preparing W as a single-parent child of da1d939e that changes result.md only. It applies C1–C3 there, leaves derivation files and gate-table.json as executed, and re-reads the six A4 heads through MCP. An independent exact-head review of W will recheck C1–C3. Not authorized: E, reconciliation, receipt, merge.

---

Generated by Claude Code

[amaybaum]

Independent exact-head review of the S12 halt commit W = 814e57f985781e7d88a57d85526c21aa5f6c6526: ACCEPT WITH CHANGES (all non-material)

Commissioned under #1031 c.6659415927 (part B, "Obtain an independent exact-head review of W"). The reviewer is independent, model-authored and read-only, and did not write W.

1. Structure and S12

• W's only parent is da1d939e. The chain F → 3efb9233 → 291925da → da1d939e → W is linear and single-parent (G5).
• da1d939e..W changes only result.md, mode 100644, +121/−18. Every path changed from F to W lies in the record directory (G6).
• The protected-text hash is a93f3f33…bb63. preregistration.md (blob c7505385) and amendments/ (tree 0debf636) are identical to F (G7). derivations/ (tree dc9dc56d) and gate-table.json (blob ea526d9f) are identical to da1d939e.
• S12 is met: W is the single-parent child of the last execution commit; the withdrawal invariant holds vacuously, because the governed paths at F are record entries only; result.md is present; no control-plane change after F; no E.

2. Recheck of C1–C4 (review c.6659587647)

• C1, §2: correct. Single-subtype row tokens (§5.4); A2's set belongs to the class token; G2 decides (a) first; CC-IMP's fixture puts δ_0 on E's cells, so T1 fires by construction.
• C2, §3: correct against line 56 and G5. With w_certified: false and hsel_content: [] in all three rows, Λ = the hypotheses field plus H-law, without H-sel. No token carries Λ.
• C3, §4: applied. The P1 grounds rest only on the quoted c.6658964414 and c.6659415927. c.6658841280 survives only in §1.1, lines 17 and 30, which W did not change: once inside the owner's own quoted text and once as a provenance pointer.
• C4: applied in result.md (lines 130–131), not in the derivation file. Acceptable.
• No new value, token or outcome.

3. Halt contents

• "halted" appears at lines 3 and 295. The reason is recorded verbatim (lines 174–175, 296).
• The reviewer re-read the six PR heads via MCP. All match §8.3 and A4 at F: Research (WIP): SUBSTATE-1 — internal sub-states and visible recurrence #1009 d4a80fc8, Research (draft): MR-1 — measurement as physical record formation, preregistration only #1023 af1e9be2, research: SUBSTATE-2 recursive inverse-limit substructure and depth-feedback controls #1011 947761dd, research: OI ↔ constructive QCD bridge and gauge-action non-uniqueness #1003 0efd67a6, Introduction paper: reconcile sources and integrate scoped theorem recoveries #918 44236326, Research (draft): RR-1 — ready-resource sourcing, native-Y public-port interface #1025 e02544d5.
• All candidate heads are listed with full, resolving ids.
• The quotes of c.6659415927 and c.6658964414 are exact word for word, but see W2.
• S0/P0 outcomes are unchanged. P1 and P3 are not evaluated; no E; class_tokens is {}.

4. Overstatement: none. The P0 ABSENT tokens are scoped to (π_0, branch, access, instance), "not a general impossibility result".

5. §A.42 item 8

• The K1-RDY: paper-only readiness investigation under fixed selection rule #1031 comments: provenance, with the text quoted with author and time.
• PR K1-RDY-1: native V3 round, halted after P0 (paper-only; S12 withdrawal and halted receipt) #1126 (the earlier review): provenance for the halt's precondition. The outcomes rest on the in-record derivations.
• F and the execution-commit ids: support on main once the round lands, as ancestors of Q.
• No unclosed gap.

6. Register: no local paths, closing keywords or secrets.

Requested changes (non-material, result.md)

• W1, lines 159–160: two line citations are off by one: P0-B-eq-d1.md should be 60–61 and P0-B-ct-d1.md should be 59–60. P0-B-pl-d1.md 64–65 is correct.
• W2, between lines 226 and 228: mark the elision with > […]. Part A and the paragraphs before part B are omitted, but the quote reads as continuous.
• W3, line 248: cite the review exactly, as "PR K1-RDY-1: native V3 round, halted after P0 (paper-only; S12 withdrawal and halted receipt) #1126 comment 6659587647".

Applying any of these needs a new single-parent child of W, which then becomes the withdrawal commit, with the current W counting as an execution commit. That needs owner allowance.

Observations (no change requested)

• O1: the §8.3 A4 reads (13:46:10Z) preceded this PR's posting of the earlier review (13:46:27Z). Part B is conditioned on the review returning, not on its posting.
• O2, for the receipt: architecture's T5 is ambiguous about whether W is listed in candidates. F has its own f field and should not be listed as a candidate, even though §8.4's table includes F, labelled F.
• O3: line 130 says POS-X's G5 is "not run", while derivations/S0-POS-X.md lines 73–81 still assembles Λ by G5's rule as a cross-check. No outcome is affected.

Next: the exact W SHA and the reconciliation proposal go to the owner for approval (c.6659415927, part C). Not authorized: reconciliation, receipt, merge, E.

---

Generated by Claude Code

[amaybaum]

Independent exact-head recheck of W′ = 1280b76edb6d4a2493f43fa7d5c797bdfd6fb0fb: ACCEPT

This recheck is the one required by the owner's ruling at #1031 c.6660591343, item 2. It was done by the same independent reviewer as c.6659744122, read-only.

Structure.

• W′'s single parent is W 814e57f9.
• The chain F → 3efb9233 → 291925da → da1d939e → W → W′ is linear and single-parent.
• git diff 814e57f9 1280b76e touches result.md only, mode 100644.
• The protected hash is a93f3f33…bb63.
• preregistration.md (c7505385), amendments/ (0debf636), derivations/ (dc9dc56d) and gate-table.json (ea526d9f) are identical to W.
• S12 holds with W′ as the withdrawal commit and W as execution commit 4. The withdrawal invariant holds vacuously, because the governed paths are record entries only.

Corrections.

• W1: P0-B-eq-d1.md 60–61 and P0-B-ct-d1.md 59–60 match the files.
• W2: > […] now marks the elision before part B.
• W3: the citation reads "PR K1-RDY-1: native V3 round, halted after P0 (paper-only; S12 withdrawal and halted receipt) #1126 comment 6659587647".

§8.4. W is listed as execution commit 4, "last execution commit and parent of W′", and the da1d939e row reads "parent of W". Both are consistent with G5 and S12.

§8.6. The quote of #1031 c.6660591343 item 2 is exact word for word through "Require independent exact-head recheck.", with the elision after it marked. Author and times match.

Nothing else changed: no outcome, value, token or statement.

§A.42 item 8.

• K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6660591343: provenance, quoted with author and time.
• PR K1-RDY-1: native V3 round, halted after P0 (paper-only; S12 withdrawal and halted receipt) #1126 c.6659744122: provenance.
• 814e57f9: support on main once the round lands.
• No gap.

Next, per the ruling: reconciliation, the halted receipt Q and tools/v3_verifier.py --verify-round Q, which must print VERDICT HOLDS. Not authorized: E, or the merge of #1126.

---

Generated by Claude Code

[amaybaum]

K1-RDY-1 halted landing prepared: reconciliation Λ, receipt Q, VERDICT HOLDS

Authority: owner ruling #1031 c.6660591343, item 2: "authorize preparation of reconciliation, the halted receipt and --verify-round Q verification. Require VERDICT HOLDS." The sole branch committer is writer-k1rdy1-exec.

Λ = 2593e382ddc9d8a6ac2896744ef2e2d95612876b. Its first parent is main 9f9dbc28 (after #1134) and its second parent is W′ 1280b76e.

• D 4b4b437c lies on main's first-parent chain, and the merge had no conflicts.
• git diff 9f9dbc28 2593e382 is byte-identical to git diff 4b4b437c 1280b76e: 28 files, all in the round's record directory. The landing therefore adds exactly the round's own diff (§A.39 step 3). The coordinator verified this independently.

Q = 7503813d6e88d6889eb1366fada67f26ecd25fe1. Its single parent is Λ, and it adds only verification/receipts/K1-RDY-1.json. The round is non-sealing, so there is no seal record. The receipt, built with tools/v3_receipt.py --status halted, records:

• Identity: status: halted, kind: non-sealing; d 4b4b437c, f d9b3de01, withdrawal W′ 1280b76e.
• Control plane: 12 control-plane blobs; governed-paths digest c7ecedf3…0d1d.
• candidates: 3efb9233, 291925da, da1d939e and W 814e57f9, each with its tree. F is not listed (reviewer note O2). The verifier requires W, as the last execution commit, to be a candidate (s12:last-execution-commit-not-a-candidate).
• landing: base 9f9dbc28, reconciliation Λ, no resolved paths.
• attestations: F's owner designation (K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6658423367, amaybaum, 12:50:29Z) and F's check-run (verify run 42026370642: workflow_dispatch, head_sha = F, 40 of 40 jobs successful, recounted by the coordinator via the API). There is no E attestation.
• absent: e, tree_e and execution_delta_digest, each halted-before-certification.

Verification was run in a fresh full-history clone from GitHub (not shallow, no alternates), with the verifier unmodified:

• tools/v3_verifier.py --verify-round 7503813d → VERDICT HOLDS (exit 0). As always for host records, the attestation lines read "recorded, unverified".
• --receipts → RECEIPTS 53 receipt(s), all hold.
• legacy_records_check.py → LEGACY 303 record(s) in 75 closed namespace(s), all intact.
• v3_receipt.py --self-test → OK.

Not done: E designation, dispatch, merge. The ordinary PR CI at Q runs on this PR. Merging #1126 needs a separate owner decision.

---

Generated by Claude Code

[amaybaum]

Release gate fails at W and at Q: a repeated paragraph in result.md (duplicate_check)

Runs. PR verify run 42047242764 at Q 7503813d failed in the job "Mathlib bridge", step "Release gate", with FAIL duplicate duplicate_check: FAILED (1 paragraph(s) repeated within a single file). The same failure occurred at W 814e57f9, run 42039121271. The run at W′ (42046291348) failed earlier, in Build and in "Report toolchain"; those are transient infrastructure failures and say nothing about content.

Cause. verification/programmes/oi-qm/reconstruction/round-k1-rdy-1/result.md at W, and therefore at W′, Λ and Q, contains the same A4 readings table twice: once in §1.5 and once in §8.3. The two blocks are byte-identical because all six readings were identical. The table is at least 200 characters long, so tools/duplicate_check.py counts it as a repeated paragraph. At da1d939e the file had no repeat; W introduced the second copy.

Correction of earlier reports.

• The W writer reported duplicate_check OK. The coordinator's own "local reproduction" at Q also reported OK, but it ran on an extraction that did not contain the record directory, so it was invalid.
• Running the checker's paragraph rule directly on result.md at W and at Q reproduces the repeat.
• --verify-round Q still prints VERDICT HOLDS, because the verifier does not run the release gate. The receipt is protocol-valid, but the PR is not green.

Effect. #1126 cannot land with a red required check. Fixing the text requires re-cutting the withdrawal commit, because Λ and Q already sit on top of W′, and once a receipt holds, the record directory is frozen at the receipt commit (G13). The re-cut path and any branch-head movement need owner approval; it is being put to the owner. Nothing has been changed.

---

Generated by Claude Code

[amaybaum]

Re-cut under #1031 c.6661119426: W″ = 93e1fd3c2a6b231092bde89d2a1bf71f4ebfbc42, independent exact-head review ACCEPT

Preservation and branch move.

• The old Q 7503813d6e88d6889eb1366fada67f26ecd25fe1 is preserved at refs/heads/archive/k1-rdy-1-q1-20261105. It was pushed first and confirmed with git ls-remote.
• The branch was then moved to W″ with --force-with-lease, expected head 7503813d. The coordinator confirmed both refs on the remote.

W″. It is the single-parent child of W′ 1280b76e. It changes result.md only (+20/−15). §8.3 now cites the retained §1.5 A4 table instead of repeating it byte for byte, and the writer made consistency edits in §8.4, §8.6 and the new §8.7.

Review. Independent and read-only:

1. Structure. Every other file is identical to W′, and the protected hash is a93f3f33…bb63. The chain F → 3efb9233 → 291925da → da1d939e → W → W′ → W″ is linear and single-parent. S12 holds with W″ as the withdrawal commit and W, W′ as execution commits 4 and 5.
2. Duplicate rule, applied to the file contents with duplicate_check.py's rule (#{2,4} headings, and paragraphs of at least 200 characters). As a control, it flags W′. At W″ nothing repeats in result.md or in any other .md file of the record directory. The coordinator reproduced this independently.
3. §8.3 keeps the re-read facts: the 13:46:10Z–13:46:12Z window, the MCP get instrument, and that all six heads equal the retained table.
4. Consistency edits are faithful. The §8.7 quote matches c.6661119426 word for word, with the elision marked. The comment time is 14:54:39Z, and W″ was committed at 14:56:07Z. No outcome, value, token or reading changed.
5. §A.42 item 8: no gap. The record does not reference the archive ref or the old Q.

One optional wording point is not applied: §8.3 says the §1.5 table records "the readings of §§1.5–1.6", but the §1.6 readings are given in §1.6's text.

Next. Rebuild the reconciliation and receipt against current main, then --verify-round (must print VERDICT HOLDS), the repository checks including duplicate_check, and green PR CI. Merge is not authorized.

---

Generated by Claude Code

[amaybaum]

K1-RDY-1 corrective landing complete. Ready for the owner's separate merge decision at Q′ = aa2eefdc0270381fdf5bbaa59d54e99660c27394

Authority: #1031 c.6661119426 (the re-cut); #1031 c.6660591343, item 2 (landing preparation). Merge is not authorized.

Chain. F d9b3de01 → 3efb9233 → 291925da → da1d939e → W 814e57f9 → W′ 1280b76e → W″ 93e1fd3c (withdrawal) → Λ′ 776593f2 → Q′ aa2eefdc. The old Q 7503813d is preserved at refs/heads/archive/k1-rdy-1-q1-20261105.

Reconciliation Λ′ 776593f2d3d4a2a27e8b7b67c870b3b1ccdc3215.

• First parent is main 9f9dbc28; second parent is W″.
• D lies on main's first-parent chain, and the merge has no conflicts.
• git diff 9f9dbc28 776593f2 is byte-identical to git diff 4b4b437c 93e1fd3c (28 files, all in the record directory). The coordinator verified this independently.

Receipt Q′. It is the single-parent child of Λ′ and adds only verification/receipts/K1-RDY-1.json.

• status: halted, withdrawal W″.
• candidates: 3efb9233, 291925da, da1d939e, W, W′. F is not listed.
• F's owner-designation and check-run attestations; no E.

Verification. All tools were run unmodified at Q′ in a fresh full-history clone (not shallow, no alternates):

• --verify-round → VERDICT HOLDS
• --receipts → RECEIPTS 53 receipt(s), all hold
• legacy_records_check → 303 records intact
• every release-gate check that does not need Lean passes, including duplicate_check: OK

PR CI at Q′: verify run 42050323384 (pull_request) succeeded: 25 jobs passed, 1 matrix template skipped. That run includes the full release gate with lean-axioms.

Scientific outcome (unchanged):

• S0: all 11 controls passed.
• P0: B-pl, B-eq and B-ct each ABSENT, scoped to (π_0, branch, access, instance).
• P1 and P3 not evaluated.
• No E designated and no class token issued.

---

Generated by Claude Code

[amaybaum]

Owner decision: merge of #1126 approved, subject to final integration checks

amaybaum, Claude Code session chat, 2026-11-05 (about 15:15Z):

#1126 is approved for merge, subject to final integration checks against the updated main. […] The merge accepts the verified record of a halted research round, not a completed P1 investigation or a new physical theorem. […]

Merge instructions

• Finish Retain the 28-module Introduction axiom-report extension (blob 97a2377d) as unreviewed evidence (#941) #1132's main CI and complete AGENTS.md §A.45: the project board tracks Issues only #1136's already-approved integration and merge.
• Recheck K1-RDY-1: native V3 round, halted after P0 (paper-only; S12 withdrawal and halted receipt) #1126 against the resulting main, confirming that its scientific content, receipt and protected records are unchanged.
• Confirm the integration passes required checks, including VERDICT HOLDS, duplicate_check and CI.
• Merge K1-RDY-1: native V3 round, halted after P0 (paper-only; S12 withdrawal and halted receipt) #1126 using a normal merge commit, with no squash or rebase, provided the approved Q′ head remains unchanged.
• Verify post-merge main CI and post the merge record.

If integration requires changing Q′ or the receipt, return for a new exact-head decision.

Keep #1031 Active. Completing K1-RDY-1 does not close the parent programme; K1-P1-1 (#1135) remains its next research stage, with code drafting, F designation and execution requiring separate authorizations.

Integration status. After #1132 landed (b28888ff), GitHub reports this PR's mergeable_state as behind. The repository requires a branch to be up to date before merging; #1129's first merge attempt was refused for that reason. Q′ aa2eefdc therefore cannot merge unchanged once main has moved.

Under §A.39 step 3, bringing the round up to date takes a further reconciliation Λ₂ (first parent: the then-current main; second parent: the previous receipt commit Q′) and a new receipt commit Q₂ that records the new landing. That changes the head and the receipt's landing field. Per the owner's condition, the new exact head will come back for a new decision.

Plan.

1. Land AGENTS.md §A.45: the project board tracks Issues only #1136 first, so that main moves only once more.
2. Prepare Λ₂ and Q₂ under the existing landing-preparation authorization (K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6660591343, item 2).
3. Confirm the scientific content and protected records are unchanged, and that --verify-round prints VERDICT HOLDS. Also confirm duplicate_check and green PR CI on the exact head.
4. Return with that head for the owner's decision.

---

Generated by Claude Code

[amaybaum]

Second reconciliation complete. New exact head Q₂ = 6664954715b477c84a00684c83fa218d93e7c6a2 for the owner's decision

The owner's approval (c.6661612166) was conditional on Q′ staying unchanged: "If integration requires changing Q′ or the receipt, return for a new exact-head decision." main moved (#1132, #1136), and the repository requires an up-to-date branch. A second reconciliation was therefore built, and it changes both the head and the receipt's landing field.

Λ₂ = 8ac011df62e69f802939e644dabe9fa8f9c97a8f

• First parent: main 63b1cd45. Second parent: Q′ aa2eefdc.
• D 4b4b437c and Λ′'s base 9f9dbc28 both lie on main's first-parent chain.
• No conflicts.
• The record directory at Λ₂ is the same tree as at W″ (3f11c562). The coordinator verified this independently.
• git diff 63b1cd45 66649547: 29 files. These are the 28 record files, byte-identical to diff(D, W″), plus verification/receipts/K1-RDY-1.json.

Q₂

• Single parent: Λ₂.
• Changes only verification/receipts/K1-RDY-1.json, and only its landing section: base 63b1cd45, with reconciliations = [Λ′ 776593f2, Λ₂ 8ac011df] in that order.
• Unchanged from Q′: d, f, withdrawal W″, candidates (F not listed), the attestations (F only, no E) and absent.

Verification. All tools ran unmodified in a fresh full-history clone:

• --verify-round Q₂ → VERDICT HOLDS
• --receipts → 53 receipts, all hold
• legacy records → 303 intact
• every non-Lean release-gate check passes, including duplicate_check: OK
• result.md, gate-table.json, derivations/, preregistration.md and amendments/ are identical to W″
• the protected hash a93f3f33…bb63 holds

PR CI at Q₂: verify run 42057379997 (pull_request) succeeded, 25 jobs green and 1 matrix template skipped. That includes the full release gate with lean-axioms. main is still 63b1cd45.

Scientific outcome, unchanged: S0 11/11 pass; P0 B-pl, B-eq and B-ct are all ABSENT (scoped); P1 and P3 were not evaluated; no E was designated; no class token was issued.

Merge is not authorized. This awaits the owner's exact-head decision at Q₂.

---

Generated by Claude Code

[amaybaum]

Merge record: K1-RDY-1 landed (owner approval #1031 c.6662764727)

Merged at the approved exact head Q₂ 6664954715b477c84a00684c83fa218d93e7c6a2 as the normal merge commit a8ab8dae3e9af7376c9233f4d474d7ccec1ecdb2, with parents 63b1cd45 (main) and 66649547. There was no squash and no rebase, so the round's commit chain is intact on main.

Before the merge: the head was 66649547 and main was 63b1cd45, both unchanged since the decision.

After the merge:

• Tree: the merged tree equals the tree of Q₂.
• main CI: verify run 42061118983 at a8ab8dae succeeded: 25 jobs green and 1 matrix template skipped. That run includes the release gate's v3-receipts step, which verifies every receipt on main, among them verification/receipts/K1-RDY-1.json.
• Receipt verification: tools/v3_verifier.py --verify-round 66649547 printed VERDICT HOLDS in a fresh full-history clone (c.6662407014).
• Refs: the old Q 7503813d is preserved at refs/heads/archive/k1-rdy-1-q1-20261105. The round branch was removed on merge; its commits are reachable from main.

Scope. This lands the verified record of a halted round: S0 passed all 11 checks; P0 is ABSENT for B-pl, B-eq and B-ct, each scoped; P1 and P3 were not evaluated; no E was designated and no class token was issued. It is not a completed P1 investigation and not a new theorem. #1031 stays Active for K1-P1-1 (#1135).

---

Generated by Claude Code