[amaybaum]

WIP / UNVERIFIED. This is the guiding draft PR for #1049: what selects linear (A5) over affine microscopic dynamics?

• Authority: paper-only (owner, chat 2026-11-04; Linear vs affine microscopic dynamics: what selects A5? #1049 c.6622105211). On 2026-11-08 the owner, in chat, authorized the corrective commits and a fresh independent review, but not a merge, computation or any strengthening.
• Writer: the coordinating session.
• Base: main at 1609dc440f5a12d05b750f6d894bb91d09a099a3.
• Head: 0e7e412a42401616d5c2f1829a20b78becb828a4. PR CI run 42458082294 is green: 25 success, 1 skipped placeholder. The branch merges cleanly into current main.

Landing condition

#1083 lands only by a merge that keeps 8de75f7c and every later commit through the landed head reachable from main. Squash or rebase is not allowed. The Q4 note's claim "§1 as frozen at 8de75f7c" and its revision record depend on those commits (§A.42 item 8; fresh review SF5/D1).

Files and reviews

All reviews are independent model reviews by subagents, not human. The texts are posted verbatim; reviewer identity is unnamed and times are unattested.

• A5-LINEAR-AFFINE.md (core).
  • Reviews: 4e096997 ACCEPT WITH CHANGES; fc5cf7e9 ACCEPT WITH CHANGES (texts, c.6718252075); f07f55d0 ACCEPT (c.6622660127).
  • Later corrections, reviewed below:
    • records the owner's Q1/Q2 decisions (Linear vs affine microscopic dynamics: what selects A5? #1049 c.6624263637);
    • scopes the D3 row and §6 Q4 to the (A-lin) reading;
    • rests C2 on SM:240;
    • narrows C3 to B0.3(e), labelled as an internal check.
• A5-Q4-STAGE2D.md (Q4 note).
  • Original reviews of 8de75f7c, c155200d, a63b7fda and 645bae17 (ACCEPT): texts, c.6718263472.
  • Fresh independent review at 22ad8d69, requested by the owner: ACCEPT WITH CHANGES, with 0 blocking findings, 6 should-fix and 7 nits (text, c.6718607630). It re-derived the mathematics by hand and found no errors.
  • Applied in 0dd2d5ad, ae3925c5 and 0e7e412a. Delta check: ACCEPT. Confirmation at 0e7e412a: ACCEPT (text, c.6718644099).
  • §1, the preregistered rule, is byte-identical to 8de75f7c.

Findings (WIP / UNVERIFIED)

• Fixed-point criterion. An affine bijection relabels to a linear one by a time-independent change of coordinates exactly when it has a fixed point. S1 has none.
• Shift gauge. Under the corpus's uniform-shift gauge (SM:96, Substratum:270), A5 fixes a gauge whenever κ≢2.
• Criterion 2. Shift-invariant readouts give identical multi-time statistics under the affine and linear rules. Absolute readouts agree only up to a time-dependent relabelling of outcomes.
• Criterion 3. None of the eleven candidate principles selects linearity on its own.
• Owner decision Q1. Main's absolute interface is the formal baseline; shift-invariant observation is a conditional alternative.
• Owner decision Q2. Amplitude shifts are mathematical relabellings unless their physical gauge status is justified.
• Q4 outcome. (iii), for O1–O4 only, resting on the (A-lin) reading. It holds given H-b, on the lift, for a uniform offset. O5 is not examined. A5 keeps its status as a posit.

Open, and not authorized

• Q3, the A6 offset extension, O5 and the §7 computation.
• The §5 proposals to the parent records: D16, R10, D3(iii), and naming the route in D3.
• Integration is a separate owner decision. It requires the landing condition above.

🤖 Generated with Claude Code

https://claude.ai/code/session_013T6pfKuv5mi3MXtKjH2tZb

[amaybaum]

Independent review of checkpoint 1 (4e096997): ACCEPT WITH CHANGES. 0 blocking, 5 should-fix and 7 nits.

Confirmed by hand

• Φ^t = τ_{p(t)}L^t.
• Prop B0.2, the fixed-point characterization of linearizability, with both directions proved (§A.34). "Relabelling" ranges over all bijections.
• The S1 analysis, including the q∣N control.
• Theorems 2.1 and 2.2 and Corollary 2.3, with the coordinate/gauge distinction kept.
• The kernel-consumer inventory.
• Provenance: the witness entered SM:240 through H2 (MI-1 H2: center independence gives an affine rule, not a linear one #1050), not through S1.
• A5's status is unchanged, and no finite failure is stated as a no-go.

Should-fix

• Stage 2(d). It may use b = 0 implicitly, because for b ≠ 0 the commutant of the affine rule loses the singlet U(1). This is a candidate hidden use of A5. Q4 stays open.
• The readout class 𝒜_W is misdescribed. For S1 it is the commutant of the whole uniform two-slice group, so it is not "every function mod a constant". The theorem itself is unaffected.
• §5(b) contradicts the file's own gauge table. Under the declared uniform-shift gauge the restatement is b ∈ (κ−2)ℤ/q, not "has a fixed point".
• Parity scoping (N8/N9).
• N9's "A6 relocates A5" depends on the drafter's chosen extension.

Next. The drafter revises, paper-only with no scripts, and the revision then gets a delta review.

---

Generated by Claude Code

[amaybaum]

Checkpoint 2 (WIP / UNVERIFIED): fc5cf7e99b0eda14820761a0996792f20eccec78. It applies all 12 findings of the review of 4e096997, in one file (+201/−107).

• Stage 2(d). It may use b = 0 without saying so. This is recorded as a candidate hidden use of A5, carried into D16, R10 and §5(d). Q4 stays open.
• Readout class. It is restated through the uniform two-slice group C. Theorem 2.1 is stated for a general invariant Y. The velocity is excluded.
• §5(b). It now matches the declared uniform-shift gauge (b ∈ (κ−2)ℤ/q) and names the gauge each form assumes.
• Scoping. The parity statements now cover even q, and the odd-q condition is stated. N9 is scoped to the extensions E1–E3. A6Cov says nothing about offsets.
• Nits are applied.

The work is paper-only. A delta review of this exact commit is running.

---

Generated by Claude Code

[amaybaum]

Checkpoint 3 (f07f55d0112dbf6f3888b824748b2fc64a79c32b): ACCEPT. The paper-only deliverable for #1049 is complete at this commit.

Review trail: 4e096997 ACCEPT WITH CHANGES → fc5cf7e9 ACCEPT WITH CHANGES → f07f55d0 ACCEPT.

• The last step applied the delta review's one should-fix and three nits in the reviewer's own wording.
• The reviewer confirmed the diff holds exactly those fixes: +5/−4.
• Every new derivation in fc5cf7e9 was redone by hand and is correct.

Reviewed result (paper, unpromoted; A5's status unchanged)

• Fixed points. An affine bijection can be relabelled to a linear one exactly when it has a fixed point, proved in both directions. S1's constant offset drives a static mode of the gapless case.
• Gauge-invariant form. Under the declared uniform-shift gauge, the condition is b ∈ (κ−2)ℤ/q. "Has a fixed point" is the gauge-invariant form only if site-dependent translations are gauge.
• Criterion 2: reached for a specified class.
  • For readouts in 𝒜_C, the commutant of the uniform two-slice group, the affine and linear rules give identical multi-time statistics.
  • At Main's absolute interface, they agree only up to a time-dependent relabelling of outcomes, and every relabelling-invariant statistic agrees.
• Criterion 3: scoped negative. None of the 11 candidate principles selects linearity without either admitting S1 or selecting only up to translation. "A6 relocates A5" holds only under the extensions E1–E3 examined.
• Criterion 1: not reached.
• Candidate hidden use of A5 (R10). It sits at Substratum Stage 2(d): for b ≠ 0, the affine rule's stabilizer loses the singlet-block U(1), which Substratum:163 identifies as hypercharge. Whether another U(1) can serve is not examined.

Owner questions (§6)

• Q1, decisive: is OI's lattice observer interface absolute (Main §1.4) or relational (𝒜_C, as Substratum:270 suggests)?
• Q2: is the amplitude-scale gauge a relabelling or a dynamical symmetry?
• Q3: which extension of A6 to offsets is intended?
• Q4: at Stage 2(d), is the stabilizer taken of the linear part or of the affine rule? This needs a referee reading.
• Q5: a citation provenance point.

Computation. §7 proposes exhaustive small-case checks of R2, R3, R6, R7 and Prop 2.6(c), for separate authorization only.

Nothing is promoted, merged or propagated.

---

Generated by Claude Code

[amaybaum]

Checkpoint 4: Q4, the Stage 2(d) stabilizer under an affine offset (WIP / UNVERIFIED).

• Commit 8de75f7c489941d32fa4bcfce28b367fdc430791 adds A5-Q4-STAGE2D.md. It is paper-only.
• Drafted outcome, not yet reviewed. It is conditional on H-b: the cubic group acts on the offset b as on the field.
  • For b ≠ 0, the rule's stabilizer is H = U(3)×U(2) ≅ G₀/U(1)_diag.
  • H/(H∩K) ≅ S(U(3)×U(2)), and Theorem 15's hypercharges are unchanged.
  • Residue: the "singlet block = hypercharge" wording holds only modulo the common phase, and 10 of SM:493's 22 cross-block modes are broken explicitly by b.
  • Side findings:
    • the papers disagree on whether M or Σ carries (3,2,1);
    • the SM:587 count of non-gauge U(1)s is unconfirmed.
• Independent exact-head review is running. Nothing is adopted or propagated.

---

Generated by Claude Code

[amaybaum]

Review provenance (1 of 2): core-file reviews of A5-LINEAR-AFFINE.md

These two review texts are preserved verbatim from the coordinating session's scratch space. Until now that was the only copy, and scratch space is discarded when the session ends. This preservation was approved as step 1 of the #1049 handoff plan, which the owner accepted in chat on 2026-11-08 ("Proceed as suggested").

What these are. Each text is an independent model review by a subagent, not a human review. The files do not name a reviewer. Each time below is the file's modification time on the session host, which nothing attests. The sha256 values are of the files exactly as stored.

Reviewed SHA	Verdict	File mtime (UTC, unattested)	sha256
4e096997	ACCEPT WITH CHANGES (0 blocking / 5 should-fix / 7 nit)	2026-11-04 04:27:13	34291f17624c6d450415e424ac368302d8df65428b47e9dcbb8c922c0b968fbd
fc5cf7e9	ACCEPT WITH CHANGES (0 / 1 / 3)	2026-11-04 04:34:10	85ead0311e64a7904881809bbf860dcbbb95dd52f13d56b722de7f7b3bde1c35

The core file's final ACCEPT at f07f55d0 is already recorded in c.6622660127. The Q4-note reviews follow in the next comment.

REVIEW-4e096997.md

# Independent review — A5-LINEAR-AFFINE.md at `4e096997`

Object: `verification/programmes/substratum/a5-linear-affine-20261104/A5-LINEAR-AFFINE.md`. HEAD was confirmed as `4e096997227a…`, a single commit on `1609dc44` that adds one file. The review is paper-only. Nothing was edited and no code was run.

**Verdict: ACCEPT WITH CHANGES.** 0 blocking, 5 should-fix, 7 nit.

The core algebra is correct: B0.1, B0.2, B0.3, Theorems 2.1 and 2.2, and B3.1–B3.4. A5's status is left unchanged, and no finite failure is stated as a no-go. The provenance point is also correct.

## Findings

1. **should-fix — :119 (D3), :517–519 (Q4).** There is a candidate use of b = 0 at Stage 2(d).
   - Translation commutation makes b uniform. Isotropy (E4) makes b a cubic-fixed vector of V_K, and the only such vector lies in the A₁ block (6 = T₁⊕E⊕A₁, SM §4.6, Substratum:163).
   - On the lift, a unitary g that commutes with the affine rule must satisfy gb = b. So for b ≠ 0, the stabilizer of the *affine* rule is U(3)×U(2), not U(3)×U(2)×U(1). The common-phase step ("amplitude-scale invariance removes only the common overall phase") then has nothing to remove, because λ carries b to λb.
   - Stage 2(d) computes the stabilizer of the linear part only. That choice is itself a hidden use of A5.
   - Fix: replace "No use of b = 0 was found" with this candidate. Propagate it to D16, and to SM:344 through D3. Keep Q4 open, with the question stated as "is the stabilizer taken of the linear part or of the affine rule?"

2. **should-fix — :332–334, :420–422.** The gloss on 𝒜_W is wrong for S1 itself.
   - β = (0,1) and Lβ = (1,2), with determinant −1 ≠ 0. So W is the whole uniform two-slice group U ≅ (ℤ/5)².
   - W ⊆ U gives 𝒜_W ⊇ 𝒜_U, where 𝒜_U consists of functions invariant under *independent* per-slice uniform shifts. It does not contain every function "modulo a common constant": the velocity x_t − x_{t−1} is invariant under a common shift but is not in 𝒜_W.
   - Item 2 (:421) also needs correcting. The commutant contains non-diagonal operators, namely the T_V(w) and the Fourier projectors. A single-site sector therefore still reads momentum-type observables; it is the configuration readouts that trivialize. The count is q^{2|V|} → q^{2|V|}/|W_V| for diagonal readouts.
   - Fix: correct both statements. Consider also stating Theorem 2.1 with 𝒜_U, which does not depend on b and covers every uniform-offset rule with a given linear part.

3. **should-fix — :499–501, §5(b).** "Has a fixed point" restates A5 gauge-invariantly only if site-dependent static translations are gauge. The file's own table (:413) and N6(1) (:247–248) say they are not declared gauge.
   - Under the declared uniform-shift gauge, the restatement is b ∈ (κ−2)ℤ/q (B0.3(c)).
   - The two conditions differ. Take κ ≡ 2 with q ∣ N, as in B0.3(e): a fixed point exists, with a non-uniform profile, but no uniform shift removes b.
   - Fix: give both forms and name the gauge each one assumes.

4. **should-fix — :262–275, :205, :280–281, :288.** Several parity statements need correcting.
   - **N8.** At every even q, the units are odd, so 1−λ is even. Then b = q/2 satisfies (1−λ)b = 0 for every unit. N8 therefore fails to force b = 0 at every even q, not only at q = 2, where it is vacuous.
   - **κ ≡ 2.** κ ≡ 2 ⇔ dα ≡ 1 holds only for odd q. At q = 2, κ ≡ 2 holds for every α, so "needs gcd(d,q) = 1" (:281) is false there.
   - **N9 at K = 1, q = 2.** AddAut is trivial, so b is unconstrained. State this case explicitly.

5. **should-fix — :283–300, N9.** The verdict depends on two choices the drafter makes in extending A6: an untransported offset and additive links.
   - A third natural extension, a transported site offset b_i ↦ g_i b_i, also restricts nothing.
   - `A6Cov` (BackgroundIndependence.lean:131) says nothing about offsets.
   - Fix: say "A6 as extended here", list the three extensions, and keep the relocation verdict scoped to that list.

6. **nit — :255.** N6 also fails (c): line 246 says it is not equivalent to b = 0. Add (c) to its failing clauses.

7. **nit — :105.** "§3.2.8" should be "§3.3.8".

8. **nit — :208–209, :495–498.** The quotation and the §5(a) proposal need tightening.
   - SM:96 says "preserve the emergent observables". The phrase "leave every observable unchanged" is Substratum:270 only.
   - SM:96 already reads nonlinear relabellings "under relabelling". §5(a) should acknowledge this and aim the scoping at Substratum:270.
   - The corpus calls x ↦ λx + c "additive-group automorphisms" although shifts are not automorphisms. #1050 routed this terminology to a corpus-consistency unit; cite that.

9. **nit — :385–386, Prop 2.5.** Over ℤ/q with q ∣ N, the uniform sector and the sum-zero sector are not complementary. Scope the "Fourier mode" statement to the lift or to gcd(N,q) = 1, or state it on G/U.

10. **nit — :126, D10.** GR:679 says a formal proof "has not been written out", so the support is a derivation sketch. Label it that way.

11. **nit — :540–541, §7.** Two of the decision rules are weaker than the claims they test.
    - R2 tests only (ii)⇔(iii), which is near-tautological. Add a cycle-type test of (i).
    - R6 tests point preparations and diagonal readouts, while Theorem 2.1 claims arbitrary ρ and arbitrary Kraus operators in 𝒜_W. Narrow the rule or say what the finite check covers.

12. **nit — :189 vs :210.** B0.3(c) writes the shift's effect as (κ−2)c and B0.4 as (2−κ)c. This is harmless, but fix one convention.

## Hand checks redone

**B0.1 and B0.2.**
- p(t+1) = Lp(t) + β, which gives Φ^t = τ_{p(t)}L^t.
- B0.2 closes as the cycle (i)⇒(ii)⇒(iii)⇒(iv)⇒(i). Every direction is proved, so §A.34 is met. Relabellings range over all bijections of G.
- The argument also covers the case where "linear" means linear for some other group law on G: an automorphism of that law fixes its identity, so a fixed point is still needed.

**S1.**
- The sum over the ring gives 0 ≡ 4, so there is no fixed point.
- The zero orbit is g = 0, 0, 1, 3, 1, 0, 0, 1, with period 5; on the lift it is t(t+1)/2.
- With the 4th root 2 ∈ 𝔽₅, the symbol of A − 2I is 0, 3, 1, 3.
- Ker Δ = ⟨𝟙⟩ and Im Δ = 𝟙^⊥, with 𝟙·𝟙 = 4.

**B0.3.**
- (e), q ∣ N with q odd: Ker Δ = ⟨𝟙, (i)⟩, 𝟙·𝟙 = N ≡ 0 and 𝟙·(i) = N(N−1)/2 ≡ 0. This holds only for odd q; it fails at q = 2, N = 2.
- (f): 2·4 + 1 − 2 = 7 ≡ 2, so the uniform state v = 2 is fixed.

**B0.4 and N8′.** (I−L)(c𝟙,c𝟙) = (0, (2−κ)c𝟙), and L(c𝟙,c𝟙) = (c𝟙, (κ−1)c𝟙).

**N4.** For symmetric A, Q(t+1) − Q(t) = b(S_{t+1} − S_{t−1}), and Q′ is conserved.

**N5, N7, N9.**
- N5: time reversal holds for every F.
- N7: commutation with z ↦ 2c − z holds iff 2(β − (I−L)c) = 0.
- N9: covariance with an untransported b requires g_i b_i = b_i. The affine-link version holds identically for additive g.

**Theorems 2.1 and 2.2.**
- The induction for Theorem 2.1 checks: U_L T(w) = T(Lw)U_L, p(Δ) + L^Δp(t_j) = p(t_j+Δ), and the Kraus operators commute with T_V(w_V)⊗T_H(w_H).
- The conjugated instruments of Theorem 2.2 check.
- With Main §1.4's convention T_ij = P(j|i), T^Φ_{ij} = T^L_{i,j−p_V}.
- tr_H removes T_H.
- The Π(t)Γ^LΠ(s)⁻¹ form preserves P-divisibility and CP-divisibility.

**Corollary 2.3.** The list of relabelling-invariant statistics is correct: entropies, the conditional mutual information, the Markov property, P- and CP-divisibility, and representability up to a permutation unitary.

**Props 2.4 and 2.6.**
- On W-invariant f, f∘Φ = f∘L.
- tr U = #Fix.
- Cycle type is recoverable from the characteristic polynomial.
- 2.6(a): (0, 𝟙) at t = 1.
- 2.6(c): the coset condition.

**B3.1–B3.4.**
- B3.1: W and G∖W are disjoint, nonempty and invariant, so `not_ensembleDetermined_of_disjoint` applies. For S1, |W| = 25.
- B3.2: #cyc(L) ≤ |W|·#cyc(L̄).
- B3.4: f(nc) = n(f(c) − f(0)) + f(0).

**Sources read.**
- SM:92, 94, 96, 224, 230, 232, 240, 242, 811 and 1416 match the file's use, as do Substratum:100, 104, 159, 163, 251–262, 270, 272 and 274.
- Theorem 24's generators are time-independent.
- The cited kernel declarations exist as cited, including `Rule` at SecondOrderLayer.lean:198.

**Provenance.**
- The witness text entered SM:240 in `5008d916`, the second parent of `981a5460` (#1050).
- S1's merge `b87733a4` (#1043) does not touch SM.md.
- The #1050 description records the review over all 5^8 states and the q∣N controls, which supports C2 and C3.

**Scope and register.** §7 is a specification awaiting owner approval. The register is neutral for a verification record. A5 is left as a posit, and every negative is scoped to the candidates searched (R5).

REVIEW-fc5cf7e9.md

# Delta review: A5-LINEAR-AFFINE.md, `4e096997` → `fc5cf7e9`

**Object.** HEAD in `wt-a5` is `fc5cf7e99b0e…`. The change touches one file, with 201 lines added and 107 removed. This review was read-only and paper-only, and every check was done by hand.

**Verdict: ACCEPT WITH CHANGES.** 0 blocking, 1 should-fix, 3 nits.

## (a) The twelve earlier findings

| # | Status | Note |
|---|---|---|
| 1 Stage 2(d) | resolved, but see N1 | The four steps, D16, R10, §5(d) and Q4 are present. |
| 2 𝒜_W gloss | resolved | §3.3.1 is rebuilt on Y and C. Velocity, non-diagonal elements and counts are corrected. |
| 3 §5(b) | resolved | Both forms are given, each with the gauge it assumes. B0.3(e) carries the separating case. |
| 4 parity | resolved | N8 is corrected for even q; κ ≡ 2 ⇔ dα ≡ 1 is restricted to odd q; E1 now covers K = 1 at even q. |
| 5 N9 scoping | resolved | E1–E3 are listed, and `A6Cov` is noted as silent on offsets. |
| 6 N6 (c) | resolved | :282 |
| 7 §3.3.8 | resolved | :112 |
| 8 quotations, §5(a) | resolved | :233–234, :495–501, :567–576 |
| 9 Prop 2.5 | resolved | :445–450 |
| 10 D10 | resolved | :133 |
| 11 §7 | resolved | The cycle-type test is added at :628–630. I checked it: equal cycle types ⇔ Φ is conjugate to L ⇔ β ∈ Im(I − L). R6 is narrowed at :632–635. |
| 12 sign | resolved | The convention at :44 is followed in :213–214 and :235–237. |

## (b) Hand checks of the new derivations

- **Stage 2(d), four steps.**
  - (i) The vectors of the signed-link representation 6 = T₁⊕E⊕A₁ fixed by the cubic group are the multiples of (1,…,1), which is A₁.
  - (ii) g(Ax + b) = A(gx) + b holds for all x iff gA = Ag and gb = b.
  - (iii) A generic equivariant A is scalar on each of the three blocks, so its unitary commutant is U(3)×U(2)×U(1). The condition e^{iθ}b = b forces θ = 0, leaving U(3)×U(2).
  - (iv) e^{iθ}I commutes with A but carries b to e^{iθ}b.
  - All four steps are correct.
- **𝒜_Y and Theorem 2.1.**
  - The proof needs only p(t) ∈ Y. L-invariance of Y is not used, but it does no harm.
  - W is the smallest admissible Y, so 𝒜_W is the largest admissible class, and 𝒜_C ⊆ 𝒜_W.
  - For S1, (0,𝟙) and (𝟙,2𝟙) are independent over 𝔽₅, so W = C.
- **Counts.** C_V acts freely on G_V with |C_V| = q² whenever V ≠ ∅. The number of orbits is therefore q^{2|V|−2}, and in general q^{2|V|}/|W_V|. A single site has one orbit. All correct.
- **b ∈ (κ−2)ℤ/q.** Substituting x = y − c𝟙 gives y(t+1) = Ay(t) − y(t−1) + b + (2−κ)c𝟙. Correct.
- **B0.3(e).** At κ = 2, a uniform v gives Δv = 0 ≠ b𝟙, so every fixed point is non-uniform. An explicit fixed point is v_i ∝ b·i². It is well defined because (i+N)² − i² ≡ 0 mod q when q ∣ N. Correct.
- **E1 at K = 1, even q > 2.** Every unit λ is odd. λ = −1 gives 2b = 0, so b ∈ {0, q/2}. Conversely, (λ−1)·q/2 ≡ 0 for every odd λ. So b ∈ {0, q/2}. Correct.
- **E1 at K ≥ 2.** For any q, the map x ↦ x + x₁e₂ carries b to b + b₁e₂ ≠ b whenever b₁ ≠ 0. Correct.
- **E3.** Covariance holds identically. Correct.

## (c) New findings

**N1 — should-fix, :139 (D16).** The new text overstates the consequence. "If the stabilizer is taken of the affine rule, the hypercharge U(1) is not obtained for b ≠ 0" is stronger than steps (i)–(iv) support.
- What is lost is the **singlet-block** U(1), which Substratum:163 identifies as hypercharge.
- U(3)×U(2) still contains S(U(3)×U(2)) ≅ (SU(3)×SU(2)×U(1))/ℤ₆, which is the Standard Model group's own global form, and that contains a U(1) of its own.
- **Fix:** "the singlet-block U(1) that Substratum:163 identifies as hypercharge is lost; whether a U(1) inside U(3)×U(2) can serve instead, with the matter action and anomaly step of SM §4.9, is not examined." Align R10's "Does NOT establish" cell (:547) and Q4 (:606–608) with this. Both currently say only that the singlet is lost, which is acceptable.

**N2 — nit, :386–387.** The velocity itself is excluded from 𝒜_C, but differences of velocities between sites are included, since a per-slice constant cancels from them. A half-sentence would stop readers concluding that all velocity information is lost.

**N3 — nit, :311.** At even q > 2, κ ≡ 2 ⇔ dα ≡ 1 (mod q/2). Only q = 2 is stated.

**N4 — nit, :9.** "not re-reviewed" will be stale once this review is recorded.

## (d) Boundaries

- A5's status is unchanged at :19, :78 and :588.
- R10 is a candidate and §5(d) is a proposal.
- Criterion 1 is not reached.
- Every negative is scoped to N1–N11 and E1–E3.
- §7 remains a specification that needs owner approval.
- The register is neutral.
- C4 and the representation theorem are untouched.

---

Generated by Claude Code

[amaybaum]

Review provenance (2 of 2): reviews of the Q4 note A5-Q4-STAGE2D.md

These are the original review texts of 8de75f7c, c155200d and a63b7fda, together with the delta review and final verdict at 645bae17 (option (b) of the completion criteria in #1049 c.6666765051). They are reproduced verbatim from the coordinating session's scratch space, which until now held the only copies. Posting them is step 1 of the #1049 handoff plan, which the owner accepted in chat on 2026-11-08 ("Proceed as suggested").

What these are. Each text is an independent model review written by a subagent. None is a human review, and the files do not name a reviewer. Each time below is the file's modification time on the session host, which nothing attests. The sha256 values are of the files as stored.

Reviewed SHA	Verdict	File mtime (UTC, unattested)	sha256
8de75f7c	ACCEPT WITH CHANGES (blocking B1–B3; S1–S6)	2026-11-04 06:46:55	83da0339ee2cbb89f200213136f63f07e64d155537707cb9553e1331318e2f05
c155200d	ACCEPT WITH CHANGES (blocking D1; F1–F6)	2026-11-04 06:55:10	ca1558e9e6df6735d54bbb638e48fb30b157883a8c355e6922010f89e8b1004e
a63b7fda	ACCEPT WITH CHANGES (blocking G1; H1–H3)	2026-11-04 07:00:20	d57e6be71f46d3139cd8583f55470b21190219db55f895ee760730c4ff456ef6
645bae17	ACCEPT (whole note and the a63b7fda..645bae17 delta), 3 nits	2026-11-04 07:03:56	cd23e996e704b88d0a828260b6276e284c99931f96ef296464ffb822b2a0ef46

The three nits from the 645bae17 review have not yet been applied. Whether these unnamed, untimed reviews satisfy option (b), or a fresh review under option (a) is needed instead, is an owner decision.

REVIEW-Q4-8de75f7c.md

# Review — A5-Q4-STAGE2D.md at 8de75f7c (independent exact-head reviewer)

Target: `verification/programmes/substratum/a5-linear-affine-20261104/A5-Q4-STAGE2D.md` at
`8de75f7c489941d32fa4bcfce28b367fdc430791` (the diff against `f07f55d0` adds only this file). Line
numbers `N:` refer to the note. `SM:N`, `Substratum:N` and the others refer to `papers/*.md` or `book/*.md` at
`1609dc44`. Everything was checked by hand. A few exponent checks were done in my head and are recorded in §D.

## Verdict: ACCEPT WITH CHANGES

The central group theory is correct. The conclusion survives at group level: H surjects onto G₀/K, and the
hypercharges are unchanged. Three stated results go beyond what the note proves. One is an omitted reading,
one is a wrong group element, and one is an output that was never checked. They must be fixed before the
note is relied on.

***

## A. Blocking findings

**B1. The affine-stabilizer reading is omitted, so "strict stabilizer is H" is conditional on resonance. Affected: lines 136–145, 208–214, 221–225, 241–243, 266–268 (Q4-2, Q4-5, Q4-6, Q4-7).**
N1 restricts the symmetries to *linear* constant g. The corpus also declares uniform static shifts to be
gauge (SM:96, Substratum:270), and the parent records this as "uniform static shift … declared gauge"
(parent §3.3.7 row 2, B0.3(c)).

On the lift, suppose the uniform A₁ static mode is gapped (κ_{A₁} ≠ 2). Then the affine rule has a fixed point
p = (x*, x*), with x* = b/(2−κ_{A₁}) ∈ V₁. Every g ∈ G₀ then gives an affine symmetry g_p(z) = g(z−p)+p:
- Φ(g_p z) = gL(z−p) + Lp + β = gL(z−p) + p;
- g_p(Φz) = g(Lz + β − p) + p = gL(z−p) + p.

So the affine stabilizer is p-conjugate to G₀, not H. The general condition is this. An affine g' = (g, w)
commutes with Φ iff [g, L] = 0 and (L−I)w = (g−I)β. So the linear parts form H exactly when
β ∉ Im(I−L), i.e. at resonance.

The same omission affects other parts of the note:
- N6(d) (line 241) dismisses only the *time-dependent* g_t, citing §3.3.7 as "not established". It misses
  the static row of that same table.
- N5's tadpole remark (line 211) omits the parent's own observation that a field shift removes the tadpole
  iff the mode is massive (parent "Reading" after B0.3).
- The 10-mode residue (N6a), the fundamental-carrier residue (N4a) and the T₁ countercontrol (N4b) all
  depend on resonance in the relevant block.

**Fix.** Add a resonance hypothesis to Q4-2 and Q4-5 to Q4-7. On the lift it reads β ∉ Im(I−L), i.e.
κ_{A₁} ≡ 2. That is SM:232's observer-level normalisation α = 1/d at scalar M. It is generically false on the
block-scalar M-route. Alternatively, state explicitly that the readings count only linear internal maps.
Also add the off-resonance branch. That branch strengthens (i), but it removes the residues.

**B2. The wrong lift of the hypercharge circle is used. Affected: lines 173–174, 188–190, 265 (Q4-4).**
The element (z⁻²I₃, z³I₂, 1) is in H. But it is not a lift of Proposition 15a's circle with its
parametrisation. The note writes the target of F (S(U(3)×U(2)) ≅ G₀/K, SM:579–583) as if it were a subgroup of
G₀.

The checks:
- From SM:582, F(z⁻²I₃, z³I₂, 1) = (z⁻¹⁴I₃, z²¹I₂). That is the 7-fold cover.
- On the Hom-type carriers the note itself uses (conjugation, phase θᵢ − θⱼ), it gives charges −5, −2, +3 on
  Hom(V₂,V₃), Hom(V₁,V₃) and Hom(V₁,V₂). These are not the weights (1, −2, −3) of Q, d_R and L.
- So "with the same charges on every carrier the corpus uses for fermions" (line 190) is false for the
  element stated.

The correct H-lift is already in N2(b): (I₃, zI₂, 1) ↦ (z⁻²I₃, z³I₂). Equivalently one can use
(z⁻²I₃, z⁻³I₂, 1). F maps it to (z⁻²I₃, z³I₂), and by conjugation it gives (1, −2, −3) on those three Hom
blocks.

**Fix.** Replace the element at all three sites, and drop "S(U(3)×U(2)) = {(E,J,1)} ≤ H" as the mechanism (see S4).

**B3. Output O5 is never checked, yet the outcome covers it. Affected: lines 32, 270–274.**
O5 is the A6 promotion to local invariance. No node examines it. §6 Q4d (lines 303–304) leaves it to the
parent's open Q3/N9. Even so, the outcome says "(i) … meaning O1–O5" and "Not (ii). No output among O1–O5
fails".

**Fix.** Scope the outcome to O1–O4 and record O5 as open (parent Q3, E1–E3).

## B. Should-fix

**S1. Q4-1 and the trace verdict are overstated. Affected: lines 115–119, 262.**
Two paper sites take the symmetry group of the *dynamics* as the ambient group:
- SM:493: "The substratum dynamics is locally U(6) invariant … Σ breaks U(6) → …";
- SM:342: the "kinetic term is invariant under the custodial unitary action".

So SM's route is Stab(Σ) ∩ Sym(rule), and b enters it. The note's own N1 Σ-route (line 138) shows exactly
this. Kernel `A6Glob` is rule-level and is part of the corpus (AGENTS whole-corpus scope).

"None mentions an offset" is true only of the Stage 2(d) sites. Affine rules are discussed at
Substratum:100, Substratum:220 and SM:240.

**Fix.** Narrow the claim: "no Stage 2(d) site computes the stabilizer of an affine rule".

**S2. The `A6Glob` description is imprecise. Affected: lines 110–113.**
- `A6Glob F M` (definition at `BackgroundIndependence.lean:111`, not 107) is a proposition: every
  M-commuting g ∈ AddAut V is a symmetry of F. It is not a stabilizer.
- For F = linkF + b it does not "add the condition gb = b". It is false unless the whole commutant fixes b.
  For example, g = −id when 2b ≠ 0.
- It is stated on the finite AddAut carrier. `A6Instantiation.lean:405` excludes "the stabilizer in U(6)".
- This bears on Q4b and should be said.

**S3. The outcome label departs from the preregistered rule. Affected: lines 37–47, 270–276.**
The residues the note finds are a carrier choice (Hom-type, not V_K) and an identification valid "modulo a
subgroup". Those are, verbatim, the examples criterion (iii) lists. The hybrid label "(i) … with a scoped
(iii) residue" is not one of the rule's outcomes. Either justify it against §1, or label the carrier-level
statement (iii) and the group-level statement (i) as separate rows.

**S4. "H carries them" needs the surjection. Affected: lines 175–176.**
The SM:597 representations are representations of S(U(3)×U(2)) ≅ G₀/K. H carries them through the surjection
F|_H (N2(b)), not through the subgroup (E,J,1). Note that F(E,J,1) = (δ²E, δ⁻³J) with δ = det E, so the
subgroup is not a section.

Also, SM:583's section (E,J,det E) is not in H. The preimage in H is (E, (det E)J, 1).

**S5. The U(1) counts in Theorem 14 and SM:587 belong in the residue list. Affected: lines 230–235, 252–254.**
Under (A), H's centre torus is 2-dimensional. So SM:577's literal "Three U(1) factors exist automatically"
also holds only modulo U(1)_diag. Under (A), SM:587's "two" becomes one, even on V_K. Add both to the N6c
wording residue.

**S6. The N4a consistency check uses a placement of the note's own. Affected: lines 200–202.**
The lepton doublet on Hom(V_E, V_A₁) is the note's own illustrative placement. SM:529 places only Q_L.
"I found no conflict" should become: "one illustrative placement is consistent; no corpus assignment is tested".

## C. Nits

- Line 101: the circle z ↦ (z⁻²I₃, z³I₂) is displayed in the proof (SM:595), not in the statement (SM:593).
- Line 104: "taken as phenomenological input" is at SM:468 (Scope), not SM:464.
- Line 217: K is a 2-torus. The extra circle of the linear reading is K/U(1)_diag.
- Symbol clashes:
  - H (the stabilizer) against the Higgs representation "H = J" (line 175);
  - the offset b against Σ's block value b (line 77);
  - K = 6 against the subgroup K.
- Line 229: "Odd q" is a sufficient condition. The exact condition for losing the sign flip is 2b ≠ 0.
- Line 34: "The readings below compare them under two readings" is garbled.
- S6 could note `finrank_commutant` = 3 (`OIBridge.lean:436`) [K]. It supports Substratum:163's U(1)³
  pointwise commutant. "No module computes U(3)×U(2)×U(1)" stands.
- §A.32/§A.33: the note is exempt as a record. If any text moves into papers, remove "Does NOT" (capitals
  used for emphasis), the first person ("I found") and the self-assessment ("favour the framework", line
  236). The proposed wording "modulo the common phase, or modulo K" is register-clean.

## D. Verified correct

**Definitions and the stabilizer.**
- K at line 19 matches SM:579. From SM:582–583: F(cI₃,dI₂,t) has scalars (r⁻², r³) with r = tc⁻²d, so
  t = c²/d.
- H = U(3)×U(2)×{1} (b ≠ 0 in V₁).

**Isomorphisms and products.**
- H ∩ U(1)_diag = 1 and H·U(1)_diag = G₀. This is an internal direct product, so H ≅ G₀/U(1)_diag.
- H·K = G₀, via (A,B,z) = (A,zB,1)(I₃,z⁻¹I₂,z).
- H∩K = {(cI₃,c²I₂,1)}.
- The second isomorphism theorem gives the quotient isomorphism.
- F(I₃,tI₂,1) = (t⁻²I₃,t³I₂).

**Exponent checks.** With F exponents (4a₃−2a₂−2a₁, −6a₃+3a₂+3a₁):
- (−2,3,0) ↦ (−14,21);
- (−2,−3,0) ↦ (−2,3).

**Mode and dimension counts.**
- dim 13 versus 14.
- The 22 = 10 + 12 cross modes. b breaks exactly 11 = 36 − 25 generators: 10 cross modes plus the A₁ phase.

**Remaining checks.**
- The N3 reduction on Hom-type carriers.
- The real lift: O(1) is removed.
- Proposition 4b: the product case does not use signed permutations (SM:336).
- The V_K countercontrol is non-vacuous.
- The quotations from Substratum:102, :149, :163 and :270; Structure:152 and :1153; Explainer:830–840;
  ch02:102 and FULL:745 (identical); ch04:118; ch05:119–123; SM:340–344, :376, :380, :382, :491, :493,
  :577, :579, :585, :587 and :589–597 are accurate.
- `A6Instantiation.lean:64–67` says "neither consumed nor judged".

## E. Side findings

**N7a (whether M or Σ carries (3,2,1)): confirmed, with additions.**
- Substratum:149 and :163, Structure:152 and Explainer:838 make M block-scalar.
- SM:376, SM:493 and SM:683 ("Since the coupling matrix is scalar, M = μ·I₆ (Theorems 6–7 jointly …)") make
  M scalar.
- SM is internally unsettled. SM:376's "(Remark below)" points to SM:382, which withdraws the speed argument
  that selected the scalar M.
- Explainer:828 also lists "linearity" among the selection criteria for the M-commutant route.

**N7b (SM:587's two non-gauge U(1)s): confirmed in substance, and I would raise it from low to moderate-high
confidence.**
- The count does not depend on any inner product. On Hom-type carriers the abelian part acts through
  T³/U(1)_diag, which is 2-dimensional. Removing hypercharge leaves one direction.
- SM:587's "orthogonal" presupposes an inner product it does not state.
- Its anomaly argument cannot apply to the common phase, which acts trivially on Hom-type carriers.
- On V_K under reading (L), two directions act.

REVIEW-Q4-c155200d.md

# Delta review — A5-Q4-STAGE2D.md, 8de75f7c..c155200d

This is the independent exact-head review, done read-only with hand algebra only. Line numbers refer to the
note at `c155200d`. `SM:N` and the other paper references are to `1609dc44`.

## Verdict: ACCEPT WITH CHANGES

The mathematics in B1–B3 and S1–S6 is resolved correctly. I uphold the author's refinement to B2. One
**blocking** provenance defect is new in this revision: §1's "fixed before §3" decision rule was edited.

## Blocking

**D1. The preregistered decision rule was changed after the results without disclosure (lines 41–71).**
The section still reads "fixed before §3". The diff against 8de75f7c nevertheless changes four things:
- (i): "under reading (A)" becomes "under **every** offset reading";
- (iii)'s examples gain "or **affine** frames";
- (ii): "subgroup of H" becomes "subgroup of the offset stabilizer";
- the sentence fixing the outputs as "those the corpus derives from Stage 2(d)" is deleted.

§8 does not record any of these changes, which violates §A.29: preregister the rule and generate the verdict from it.

The amendments are conservative: they move the outcome away from (i). Under the original wording, the
(A-lin) branch already gives (iii), through the carrier choice on V_K and the "modulo a subgroup" clause. So the
label survives. The defect is provenance, not the verdict.

**Fix.** Restore the original §1 text and add the change as a dated amendment, or disclose it in §8. State
the label under the original rule as well.

## Checks requested

**1. B1, re-derived.**
- Commutation with g'(z) = gz + w:
  - Φ(g'z) = Lgz + Lw + β;
  - g'(Φz) = gLz + gβ + w;
  - so they commute iff Lg = gL and (L−I)w = (g−I)β. ✓
- The rule at SM:314 gives the uniform-mode equation (1/d)·M·2d = 2M. So κ_a = 2μ_a. ✓
- The fixed point is x* = v/(2−2μ_{A₁}). ✓
- Resonance iff μ_{A₁} = 1, iff the k = 0 mode is massless. This is Theorem 4, stated at SM:320–321 (the
  note cites SM:318, which is the rule). ✓
- At resonance, Ker(L−I) on uniform data is {(c,c) : (A−2)c = 0}, and the k = 0 V₁ part of Im(2I−A) vanishes.
  So z_g = 1. ✓
- The off-resonance conjugate is right: (L−I)(I−g)p = (g−I)β. ✓

**2. B2 rebuttal: upheld.** The charges on Hom(V₂,V₃), Hom(V₁,V₃) and Hom(V₁,V₂) are:
- (I₃,zI₂,1): (−1, 0, 1);
- (z⁻²I₃,z⁻³I₂,1): (1, −2, −3).

Their ratio is (z⁻²,z⁻⁴,1) ∈ G_v∩𝒦, which acts by (2,−2,−4), as SM:585 predicts.

My "equivalently" held only at the level of F. The lift families agree, and I checked this by hand. The
(L) family is (s, −1−s, −1−2s), from a₁ + a₂ − 2a₃ = 1. The G_v family is (−1−a, a, 1+2a). The two match
under s = −1−a. ✓

**3. The (iii) label.** It follows from the amended rule. It also follows from the original rule (see D1).

## Should-fix

- **F1 (lines 112, 203, 225–227 against line 24).** The revision uses β both for the two-slice offset and
  for the L-coefficient in Q = αB + βL + γS. Rename one of them.
- **F2 (lines 213–215, Q4-4).** The u_R/e_R claim is correct but overattributed.
  - The hand check (charges on End(V_K) at s = 1 are 0, ±1, ±2 and ±3) supports it, not SM:597. SM:597 gives
    only the determinant dressings.
  - The claim is strictly true for every lift:
    - the singlet blocks are neutral, so e_R has no home under any lift;
    - the (3,1) slots carry only ±(1+s), and u_R and d_R need magnitudes in the ratio 4 : 2, so they cannot
      both be housed.
  - Label it [D] and state that argument.
- **F3 (Q4-2, line 300; Q4-5).** At resonance, the (A-aff) stabilizer is G_v ⋉ {uniform V₁ translations}.
  Off resonance, add the translations in any resonant blocks. Say "linear parts are G_v" and "the unitary
  part has dimension 13".
- **F4 (lines 315–316).** "Taken alone, these statements meet (i)'s requirements" is not accurate. (i) also
  has a carrier clause. Say "they meet (i)'s group and charge clauses".
- **F5 (line 33).** "[O] Q2" cites an owner decision. The parent records Q2 as an open question, and no [O]
  citation appears there. Cite the decision or relabel it.
- **F6 (lines 269–270).** "Affine internal maps, which the corpus's U(K) does not state": SM:96 and
  Substratum:270 do declare shifts to be gauge. Narrow this to "Stage 2(d) does not include".

## Nits

- Line 37: cite SM:320–321 instead of SM:318.
- Line 171: "at scalar M" should be "at M = I₆", since resonance needs μ = 1.
- Line 291: "the reviewer concurs". I suggested moderate-high; moderate is acceptable.

## Other checks

The raised confidence in N8b is supported by the inner-product-free count, and is not overstated. S1–S6 and
the earlier nits are applied correctly. O5 is properly left unclassified (N7, Q4-8).

REVIEW-Q4-a63b7fda.md

# Second delta review — A5-Q4-STAGE2D.md, c155200d..a63b7fda

This is the independent exact-head review, done read-only with hand algebra only. The one network call was a
read-only GET of the cited GitHub comment. Line references are to the note at `a63b7fda`.

## Verdict: ACCEPT WITH CHANGES

One finding is blocking: the new u_R claim is stronger than the hand count that supports it.

## Blocking

**G1. "u_R … has no home … for any lift" is false as stated (N3 [D] bullet; Q4-4 "Does not establish").**
The note's own argument fixes the charges on the two colour-triplet blocks:
- Hom(V₁,V₃) carries −1−s;
- Hom(V₃,V₁) carries 1+s.

At s = −5, Hom(V₁,V₃) carries charge 4, which is u_R's weight. So u_R does have a home, at that lift. What
the argument actually proves is three narrower things:
- **(a)** e_R has no home for any lift, because the singlet blocks are neutral;
- **(b)** no lift houses u_R and d_R together, because the magnitudes |4| and |2| cannot both equal |1+s|;
- **(c)** at s = 1, the lift that houses Q, d_R and L, u_R is absent.

Restate the bullet and the Q4-4 cell to say exactly that.

## Requested checks

**1. D1 is resolved.**
- §1 is byte-identical to `8de75f7c`. I checked this by diffing the extracted section.
- §8.2 discloses the edit and withdraws it.
- One change is not recorded. `c155200d` also changed (ii)'s quantifier, from "under (A)" to "under an offset
  reading". That made (ii) existential over the readings. A2 records only the change from "subgroup of H".
  State (ii)'s quantifier under the amended reading (should-fix H1). It has no effect on the outcome, because
  no output fails under any reading.

**2. F5: the rebuttal is upheld.**
- Comment 6624263637 is by `amaybaum`, posted 2026-11-04T06:28:45Z on #1049.
- It quotes Q2 verbatim: "Treat amplitude shifts as mathematical relabellings unless their physical gauge
  status is justified."
- It also states "These decisions answer Q1 and Q2".
- Under §A.42 item 7, the record should give the author and time.
- The comment is an agent-generated transcription of chat decisions (its footer says so), posted from the
  owner's account. This is worth noting, but it does not undermine the citation.

**3. F2 and F3, checked by hand.**
- **F2:** the arithmetic is correct, but the conclusion is overstated (see G1).
- **F3:** G_v ⋉ T₀ is correct.
  - If g ∈ G_v, then (L−I)w = 0.
  - T₀ is G_v-invariant, because g commutes with L.
  - **H2:** under (A-aff), which is defined over *constant* maps, T₀ must be restricted to the uniform
    elements of Ker(L−I). These are {(c𝟙, c𝟙) : c in a resonant block}.
  - Unrestricted, Ker(L−I) can hold staggered static solutions where μ_a = −1. Translating by them is not a
    constant internal map.

## Should-fix

- **H1:** see check 1.
- **H2:** see check 3.
- **H3 (§4 and the "Effect" paragraph of §8.2): "The two classifications do not differ."** Take §1 as frozen,
  without A3, on the (A-aff) off-resonance branch. That branch's full agreement relies on uniform shifts,
  and SM:96 and Substratum:270 do state that shifts are gauge. So whether that branch is "a choice the corpus
  does not state" is contestable. (iii) overall stands on the (A-lin) branch, whose elements are Stage 2(d)'s
  U(K). Say so, rather than claiming that the classifications are identical on every branch.

## Nits

- §8.2 "Effect": A2 bears on (ii), not on (i). A3 widens (iii). The sentence "A1–A3 move the outcome away
  from (i)" is accurate only for A1.

## Checked and fine

The following are all applied correctly:
- F1, the a_B, a_L, a_S renaming;
- F4;
- F6;
- the nits from the previous round;
- Q4-2 and Q4-5, now phrased in terms of "linear parts" and "unitary part";
- N5.

The §4 frozen-rule outcome is correctly derived for the (A-lin) branch:
- (i) is not reached, because of O5 and the carrier clause on V_K;
- (ii) is not reached;
- (iii) holds for O1–O4, with the restriction to O1–O4 disclosed as A5.

REVIEW-Q4-645bae17.md

# Third delta review and final verdict — A5-Q4-STAGE2D.md at 645bae17

This review covers the diff a63b7fda..645bae17 and the whole note at `645bae17`. It was done read-only with
hand algebra only. Line references are to the note at `645bae17`.

## Final verdict on the whole note: ACCEPT

There are no blocking findings. Every blocking and should-fix item from the four review rounds is resolved:
- 8de75f7c: B1–B3 and S1–S6;
- c155200d: D1 and F1–F6;
- a63b7fda: G1 and H1–H3.

The note is acceptable as a WIP / UNVERIFIED paper-only record. Its outcome is (iii) for O1–O4. That outcome
rests on the (A-lin) branch, takes H-b and resonance as named hypotheses, and leaves O5 unclassified.

## Items checked at this head

**§1 (D1).** §1 is byte-identical to `8de75f7c`. I re-checked this by diff.

**A2′ (H1).** The row records the existential quantifier for (ii). The Effect paragraph now assigns each
amendment its own effect, and those effects are accurate.

**F5.** The [O] quotation now carries its author, time and comment id. These match the comment I fetched in
the previous round (amaybaum, 2026-11-04T06:28:45Z).

**G1: the three claims, re-derived.** From the family relation a₁ + a₂ − 2a₃ = 1 with s = a₃ − a₂, the
charges are:
- Hom(V₁,V₃), a 3: charge −1−s;
- Hom(V₃,V₁), a 3̄: charge 1+s;
- the (1,1) slots: charge 0.

So the three claims hold:
- **(a)** e_R needs charge −6, and the (1,1) slots carry 0. ✓
- **(b)** A (3,q) field fits a (3,1) slot iff q = −1−s, and its conjugate gives the same condition. u_R
  needs s = −5 and d_R needs s = 1, so no single lift houses both. ✓
- **(c)** At s = 1 the charges are {0, ±1, ±2, ±3}, and the 3 slot carries −2. ✓

The s = −5 counterexample is now stated.

**H2.** For uniform c in block a, A c𝟙 = (1/d)·M·2d·c𝟙 = 2μ_a c𝟙. Ker(L−I) on uniform data is therefore
{(c𝟙, c𝟙) : (A−2I)c = 0}. That is the resonant blocks, and G_v preserves it. G_v ⋉ T₀ is correct. ✓

**H3.** Resolved. (iii) now rests on (A-lin). The (A-aff) off-resonance branch is recorded as a contestable
caveat under the frozen §1, and the note adds that (i) is not reached overall in any case, because O5 is
not examined.

## Nits (non-blocking)

- **Line 177 contradicts lines 185–186.** Line 177 says the off-resonance "stabilizer is τ_pG₀τ_{−p}",
  while lines 185–186 say it also contains uniform translations in resonant blocks. Say "contains" at
  line 177, or "its linear parts conjugate to G₀".
- **Line 180.** {(c𝟙,c𝟙) : c ∈ V_a, μ_a = 1} should be the *sum* over the resonant blocks a, not their union.
- **N3 [D] lead-in.** "These hold for every lift" does not fit (c), which is specific to s = 1.

---

Generated by Claude Code

[amaybaum]

Review provenance: fresh independent review at 22ad8d69

On 2026-11-08 the owner asked in chat for "a fresh independent review of the resulting Q4 note rather than treating the restored transcripts as sufficient provenance". This is that review, posted verbatim. An independent model reviewer (a subagent) wrote it. It is not a human review, and no reviewer is named. The time is the file's modification time on the session host, which nothing attests.

Reviewed SHA	Verdict	File mtime (UTC, unattested)	sha256
22ad8d69	ACCEPT WITH CHANGES (0 blocking / 6 should-fix / 7 nit)	2026-11-08 07:08:51	f733e90bd8ef1f2cb6192d2e8a4dc4b5c8869fdf578d5e142969c4c325426a0c

The findings are being applied in a follow-up commit, which will get its own independent delta check. SF5's commit-ancestry point is a landing condition: #1083 must land by a merge that keeps 8de75f7c..head reachable, not by squash or rebase.

REVIEW-FRESH-22ad8d69.md

# Fresh independent review: A5 Q4 note and correction delta at `22ad8d69`

Object: PR #1083 (Issue #1049), exact head `22ad8d69bfc7aff9f969867904835f74c3419128`, single parent
`645bae17`. Files: `verification/programmes/substratum/a5-linear-affine-20261104/A5-Q4-STAGE2D.md`
("Q4") and `.../A5-LINEAR-AFFINE.md` ("core"). Line numbers are at `22ad8d69`. `SM:N` etc. are
`papers/*.md` / `book/*.md` at `1609dc44`.

This review is read-only and paper-only. Every check is by hand, or by reading sources with `git
show` and the GitHub comments API. I read the earlier model reviews in #1083 c.6718263472 only as
context, after doing my own checks. No finding below rests on them.

## Verdict: ACCEPT WITH CHANGES

0 blocking, 6 should-fix, 7 nit.

Every mathematical step I re-derived is correct: N0–N6, N8 and the §8 lift arithmetic. §1 is
byte-identical to `8de75f7c`. Each source quotation I checked matches `1609dc44`. The three
`645bae17` nits are applied correctly. Q1 and Q2 are recorded verbatim, with the right author and
time. Q3, Q4, Q5, the A6 offset extension (N9) and §7 stay open. The delta strengthens nothing.

The changes needed are these:
- The result table and the outcome do not carry H-b, the lift and the uniform-offset condition
  into every conclusion.
- (ii) is stated without the O1–O4 scope.
- The §1 controls (C-a)–(C-c) are never reported.
- The proposed R10 restatement in §5 drops hypotheses.
- Two §A.42 item 8 points must be settled before landing.

***

## Findings

### Should-fix

**SF1. Hypotheses dropped in the result table. Q4:338 (Q4-5), Q4:339 (Q4-6).**
- Q4-5's "Hypotheses" cell reads "(A-lin), or (A-aff) at resonance". Q4-6 reads "as Q4-5". Both
  omit H-b, the lift and the uniform offset.
- Q4-2 to Q4-4 carry those conditions, and the claims depend on them. dim Lie(G_v) = 13 needs
  v ∈ V_1, which is H-b (N0).
- The note's own countercontrol shows the dependence. A T₁ offset under (A-lin) gives the linear
  stabilizer U(2)×U(2)×U(1), of dimension 9 (N4b).
- So Q4-5, as tabulated, is stated more broadly than it is proved.
- *Fix:* replace both cells with "as Q4-2; (A-lin), or (A-aff) at μ_{A₁} = 1".

**SF2. Outcome scope and hypotheses. Q4:343–368.**
- (a) Q4:350 reads "**(ii) is not reached.** No output among O1–O4 fails or changes." The rule's
  (ii) ranges over O1–O5, and O5 is not examined (Q4-8, N7). So the result supports only "(ii) is
  not reached for O1–O4".
  *Fix:* write "(ii) is not reached for O1–O4; O5 is not examined."
- (b) The heading "Outcome under §1's rule as frozen at `8de75f7c`" is followed by "(iii) is
  reached for O1–O4". The O1–O4 scope is amendment A5 (Q4:364, §8.2). Under §1 exactly as frozen,
  no single label covers O1–O5.
  - This is disclosed at :364, but only after the verdict is stated.
  - *Fix:* move the disclosure into the heading or the first bullet. For example: "Outcome under §1
    as frozen, with the O1–O4 scope of amendment A5."
- (c) The (iii) bullets (:351–357) name the (A-lin) branch. They do not say the verdict holds only
  given H-b, on the lift, and for uniform offsets.
  *Fix:* add "given H-b, on the lift, for a uniform offset v ≠ 0" to :356–357.

**SF3. The §1 controls are never reported. Q4:78–85 and §4.**
- §1 fixes three controls, (C-a) to (C-c), and two required countercontrols. §4 derives the
  verdict without reporting any of them as green.
- They are green, as far as I can check by hand:
  - (C-a): at v = 0, G_v = G_0. N1's (A-aff) case reduces to G_0 for the linear parts.
  - (C-b): N0 and N1 use Theorem 7's decomposition unchanged.
  - (C-c): N3, last bullet, and Q4-4.
  - The countercontrols are N4a, a carrier on which the readings differ, and N4b, an offset in T₁.
- The verdict should be generated over controls stated as green (§A.29).
- *Fix:* add one line to §4 listing each control and countercontrol, its status and its node.
  Under (A-aff), note that (C-a) concerns the linear parts: at v = 0 the affine group also
  contains T_0.

**SF4. The §5 proposal text drops hypotheses and misses parent sites. Q4:374–378.**
- (a) The proposed R10 text reads "U(1)_S is lost under (A-lin), or at A₁ resonance". It drops
  H-b, the lift and the uniform offset. "or at A₁ resonance" also does not say that it means
  (A-aff) at resonance; under (A-lin) the loss does not depend on resonance.
  - This is the text proposed for the parent record, so it would carry the omission there.
  - *Fix:* "Given H-b, on the lift, for a uniform offset v ≠ 0: the linear parts of the offset
    stabilizer lose U(1)_S under (A-lin), and under (A-aff) at μ_{A₁} = 1. …"
- (b) Three core sites state, without condition, that "the stabilizer of the affine rule is
  U(3)×U(2)": the D3 row (core:126), D3(iii) (core:147–149) and the §6 Q4 bullet (core:613–615).
  - Within the core's own definition (core:145, a unitary acting identically at every site) this
    is the (A-lin) reading. N1's (A-aff) off-resonance branch shows that it is not the stabilizer
    of the rule under affine maps.
  - §5 names D16, R10 and D3(iii), but not the D3 row or §6 Q4.
  - *Fix:* add those two sites to §5's list, as proposals. Applying them is the owner's decision.

**SF5. §A.42 item 8: the frozen rule and the revision record depend on commits off `main`. Q4:8–12,
:343, :428–431, :477, :490.**
- Several statements rest on the existence and order of commits on the PR branch: "§1 as frozen at
  `8de75f7c`", "restored verbatim", and the review-response sequence `8de75f7c` → `c155200d` →
  `a63b7fda` → `645bae17`. None of these commits is reachable from `main`.
- Item 8 classes a verdict that depends on commits' order or ancestry as one that uses those
  commits. Copying files to `main` does not make the commits reachable.
- *Fix:* record a landing condition on the PR. #1083 lands by a merge that keeps `8de75f7c` ..
  `22ad8d69` reachable, not by squash or rebase.
  - Otherwise, quote §1's blob or section hash at `8de75f7c` in §8.2. Then the frozen-text claim
    can be checked from `main` alone.
- Core-file items that should be settled before the same landing:
  - **C3** (core:74) is supported only by "#1050 review record". #1050's merge `981a5460` landed
    manuscript files only, so that support is off `main`.
    *Fix:* re-cite C3 to B0.3(e) (core:219–222) and narrow it to that derivation's scope, odd
    prime q ∣ N. Or land the record.
  - **C2** (core:73) "#1050's exact-head review (all 5^8 states)" is off-`main` support for the
    exhaustive check. The claim itself is proved in SM:240 on `main`.
    *Fix:* mark the review as provenance, or drop the exhaustive qualifier.

**SF6. "Fixed before §3" is attested only by the author. Q4:53, §8.2.**
- §1's heading says "(fixed before §3)". §1 and the results §3–§4 first entered the history
  together, in the one commit `8de75f7c`, which added the whole file. So the history shows that §1
  was frozen at `8de75f7c` and has not changed since. It does not show that §1 preceded the
  results.
- §1 cannot be edited.
- *Fix:* add one sentence to §8.2 (not to §1): "§1 and §§3–4 entered together at `8de75f7c`; the
  precedence in §1's heading is the author's statement, not a property of the history."

### Nits

**N1. Q4:45–46.** The text cites "SM:320–321 (Theorem 4)" for "massless at k = 0 iff μ_a = 1".
That clause is at SM:322. SM:320 is the dispersion relation and SM:321 is blank. *Fix:* cite
SM:318–322.

**N2. Q4:376 against Q4:474–475.** §8.3 says "The [O] label is kept only on that quotation
(Readings, (A-aff))". §5 also has "A5's status is unchanged, per the owner [O]". This is sourced:
c.6624263637 says "A5 keeps its status as a posit". *Fix:* cite c.6624263637 at :376, or reword
§8.3.

**N3. Q4:354.** The text uses "K" for the torus. This sentence is outside the frozen §1, and the
note reserves K for 6 (Q4:25). *Fix:* write 𝒦.

**N4. Q4:283.** The sentence reads "Only the account of 10 of the 22 cross modes changes, not the
group". It sits just after the same bullet lists the A₁ phase among the explicitly broken
generators, and the group does lose U(1)_S (N1). *Fix:* "Besides the A₁ phase (N1), only the
account of 10 of the 22 cross modes changes."

**N5. Q4:8–12.** The **Revision** paragraph names the reviews of `8de75f7c` and `c155200d` only.
*Fix:* add the `a63b7fda` and `645bae17` reviews, or point to §8.

**N6. Q4:485 (the known residue).** §8.4 H2 still displays T_0 as {(c𝟙, c𝟙) : c ∈ V_a, μ_a = 1}.
- Taken alone, this is the union form that §8.5 corrects.
- My judgement: no change is required. §8 is a response record, the body (N1, :180–182) is
  authoritative and correct, and §8.5 records the correction two paragraphs later.
- If touched, the minimal edit is to replace the set with "the uniform static solutions in the
  resonant blocks (N1)". Displaying the sum a second time is unnecessary.

**N7. Core:603 and Q4:38–39 (attribution, informational).**
- c.6624263637 was posted from the `amaybaum` account. Its footer reads "Generated by Claude Code".
  It transcribes the owner's chat decisions.
- The records say it was posted "by amaybaum" and that it "records the owner's chat decisions".
  Both statements are accurate.
- Optionally add "(posted from the owner's account)", so that a reader does not take the comment
  for the owner's own words. No change is required.

***

## What I checked and how

**Mathematics. All checked by hand at the exact head.**
- **N1 (A-aff) commutation.**
  - L(u,v) = (v, Av − u) and β = (0, v𝟙), as in the core notation (core:35–37). From these,
    Φ∘a = a∘Φ ⇔ Lg = gL and (L − I)w = (g − I)β.
  - For constant w = (c𝟙, c𝟙): (L − I)w = (0, (A − 2I)c𝟙), and A c𝟙 = 2μ_a c𝟙 in block a. So
    (2μ_a − 2)c_a = (z_g − 1)v δ_{a,A₁}.
  - Hence at μ_{A₁} = 1 the linear parts are G_v, and T_0 = ⊕_{μ_a=1} uniform V_a, a sum.
  - Off resonance:
    - p = (x*𝟙, x*𝟙) with x* = v/(2 − 2μ_{A₁}) is fixed.
    - g_p = τ_p g τ_{−p} satisfies (L − I)(I − g)p = (g − I)β.
    - So the stabilizer contains τ_pG_0τ_{−p}, together with the resonant-block translations.
    - The `645bae17` nits at :177 and :181–182 are now correct.
  - The k = 0 projection argument at :178–180 also covers non-uniform w.
- **(A-lin), both routes.** For generic block-scalar M the commutant is G_0, and adding gv = v
  gives G_v. On the Σ-route the stabilizer is Stab_{U(6)}(v) = U(5)×1, and U(5)×1 ∩ G_0 = G_v.
- **N2.**
  - From SM:582, F(C,D,t) = (t⁻²(det C)(det D)⁻¹C, t³(det C)⁻²(det D)D). Check that F(𝒦) = 1.
    For (cI, dI, c²/d), the two scalars are c⁻⁴d²·c³d⁻²·c = 1 and c⁶d⁻³·c⁻⁶·d²·d = 1.
  - The factorizations (A,B,z) = (z⁻¹A, z⁻¹B, 1)(z,z,z) and (A,B,z) = (A, zB, 1)(I, z⁻¹I, z) hold.
    So G_v∩𝒦 = {(cI, c²I, 1)}.
  - The images are F(I, tI, 1) = (t⁻²I, t³I) and F(E, J, 1) = (δ²E, δ⁻³J).
  - (E, δJ, 1) is a preimage, since both determinants equal δ.
  - F(z⁻²I, z⁻³I, 1) = (z⁻²I, z³I), the lift.
  - F(z⁻²I, z³I, 1) = (z⁻¹⁴I, z²¹I), the non-lift.
  - The ratio of the two lifts is (z⁻²I, z⁻⁴I, 1) ∈ G_v∩𝒦.
- **N3.**
  - (L) family: the element (z^α, z^β, z^{1+2α−β}) gives (s, −1−s, −1−2s) with s = α − β.
  - G_v family: (z^a, z^{1+2a}, 1) gives (−1−a, a, 1+2a). The two families are equal under
    a = −1−s.
  - At s = 1 the charges are (1, −2, −3), realized by (z⁻²I, z⁻³I, 1).
  - (a) The (1,1) slots carry 0.
  - (b) A (3,q) field fits either (3,1) slot iff q = −1−s. So u_R needs s = −5 and d_R needs
    s = 1.
  - (c) At s = 1 the charges on End(V_K) are {0, ±1, ±2, ±3}, and the 3-slot carries −2.
  - The `645bae17` lead-in nit is now correct: (a) and (b) hold for every lift, (c) for s = 1.
- **N4.**
  - The illustrative placement is internally consistent. Its sign convention, a_L = +1/2, is the
    opposite orientation to the s = 1 lift. Up to conjugating the doublet this changes nothing,
    and no claim rests on it.
  - N4b: a T₁ offset under (A-lin) gives U(2) inside U(3).
- **N5–N6.**
  - dim u(3)⊕u(2) = 13 = 8+1+3+1, one more than 12.
  - Each reading leaves one surplus circle: G_v∩𝒦 under (A-lin), 𝒦/U(1)_diag under (L).
  - 36 − 25 = 11 = 10 + 1.
  - SM:493's 22 cross generators split as 2·(6+3+2). Of these, 10 involve V_1.
  - Under G_v the abelian part is 2-dimensional, so SM:587's count drops from two to one.
- **S6, the A6Glob counterexample.** For F = linkF + v, A6Glob needs gv = v for every M-commuting
  g. g = −id commutes with M and moves v iff 2v ≠ 0.

**§1 integrity.** I extracted §1 from `8de75f7c` and from `22ad8d69` and diffed them: they are
identical. The §8.2 amendment table (A1–A5, A2′) and the Effect paragraph match their stated
effects.

**Outcome generation.**
- I applied §1's (i)–(iv) myself to the N-node results, and reach the same labels.
  - (i) fails: O5 is unexamined, and the carrier clause fails on V_K under (A-lin).
  - (ii) is not reached for O1–O4.
  - (iii) holds for O1–O4: agreement only on Hom-type carriers, or modulo U(1)_diag or 𝒦. That is
    one of the rule's listed examples.
- So (iii), scoped to O1–O4 and resting on (A-lin), is justified. SF2 covers its wording.
- O5 is left unclassified (N7, Q4-8, :364). N7's E1 remark is explicitly not counted.

**Source quotations at `1609dc44`, read with `git show`.**
- SM: 96, 114, 232, 240, 314, 318–322, 336, 340–344, 376, 380, 382, 464, 468, 491, 493, 529, 577,
  579–587, 589, 593–597, 683.
- Substratum: 100, 102, 149, 163, 220, 270, 272.
- Structure: 152, 1153.
- Explainer: 828, 830, 838, 840.
- ch02:102, FULL:745, ch04:118, ch05:119–123.
- `A6Instantiation.lean` :64–67 ("neither consumed nor judged") and :405 ("the stabilizer in
  `U(6)`").
- `OIBridge.lean:436` `finrank_commutant`.
- `BackgroundIndependence.lean:111` `A6Glob` and :131 `A6Cov`.

All matches, except the line offset in N1.

**The delta `645bae17..22ad8d69`, two files and 30+/13− lines, read in full.**
- The Q1 quote (core:600–602) and the Q2 quote (core:607–608) are verbatim against c.6624263637,
  fetched with the API.
  - Author `amaybaum`, created 2026-11-04T06:28:45Z, not edited.
  - Issue 1049.
  - The comment says "These decisions answer Q1 and Q2".
- The fragments quoted at core:479 are also verbatim. That row still says 𝒜_C "is not
  established", so nothing was strengthened.
- Status changes:
  - Q1 and Q2 change.
  - Q3 is unchanged and open.
  - Q4 stays "(open)".
  - Q5 and "Open" are unchanged.
  - N9 (the A6 offset extension) is untouched.
  - §7 is still "for separate authorization only".
  - These match the comment's "Still open" line.
- Both provenance citations resolve to #1083 comments posted from `amaybaum`. Their contents match
  their descriptions:
  - c.6718252075 holds the reviews of `4e096997` and `fc5cf7e9`, posted 2026-11-08T06:30:34Z.
  - c.6718263472 holds the reviews of `8de75f7c`, `c155200d`, `a63b7fda` and `645bae17`, posted
    2026-11-08T06:31:54Z.
- I compared the three nits against the `645bae17` review's nit list: all are applied as asked.
- §8.5 is a review-response record. That fits a verification record, and it narrates nothing in
  manuscript voice. No manuscript file changed.

**Register.**
- Both headers carry "WIP / UNVERIFIED. Paper-only". Derivations are labelled [D], and owner
  decisions [O] with their source.
- Proposals are marked as proposals.
- §A.32–§A.33 govern `papers/` prose and do not apply here. Even so, no self-assessment
  overreaches.

**§A.42 item 8: off-`main` references in the two files.** Reachability was checked with
`git merge-base --is-ancestor` against `origin/main` (`d493c173`).

| Reference | Where | Class |
|---|---|---|
| `1609dc44` (base), papers and Lean at it | both | support on `main` |
| `981a5460` (#1050 merge) | core:617 | on `main`; provenance (Q5) |
| `round-h-d-a5-status-adjudication/result.md`, `A5S-1b`, `A5S-2a`; H-D-SR `SR-3` | core:76, :79, :158 | support on `main` |
| `f07f55d0` (parent at acceptance) | Q4:15 | support off `main` until #1083 lands. It lands with the note; the core at `22ad8d69` differs from `f07f55d0` only by the Q1/Q2 records, the core:479 note and the core:9 citation, none of which touches D3, D16, R10 or Q4 |
| `8de75f7c`, `c155200d`, `a63b7fda`, `645bae17` | Q4:8–12, :343, :428–431, :477, :490 | support off `main` for "§1 as frozen" and the revision order. Gap unless the landing keeps them reachable (SF5) |
| `4e096997`, `fc5cf7e9` | core:9 | provenance (review history) |
| #1049 c.6624263637 (owner Q1/Q2, Q4 priority) | core:603; Q4:4, :38–39, :472 | the decisions are quoted in the records with author and time (§A.42 item 7), so support on `main` once landed; the comment itself is provenance |
| #1049 c.6622105211 (paper-only authorization) | core:4; Q4:5 | provenance |
| #1083 c.6718252075, c.6718263472 (review texts) | core:9; Q4:417 | provenance. No mathematical claim uses them, and both records are WIP / UNVERIFIED |
| "#1050's exact-head review (all 5^8 states)" | core:73 (C2) | support off `main` for the exhaustive layer; the claim stands on SM:240 (SF5) |
| "#1050 review record" | core:74 (C3) | support off `main`. Gap (SF5) |
| #1043 | core:78 (C7), :616 | provenance; C7 is carried by Main:726 and Substratum:100 |
| #914, #1029 | core:559, :567 | provenance (targets of proposals) |
| #1049 c.6666765051 | not in the files (only in the PR comment) | n/a |

---

Generated by Claude Code

[amaybaum]

Review provenance: delta check at ae3925c5 and confirmation at 0e7e412a

The reviewer who wrote c.6718607630 checked the correction commits that apply that review. The texts below are posted verbatim. They are an independent model review written by a subagent, not a human review. No reviewer is named, and the time is unattested.

Range	Verdict	sha256 of the text below
22ad8d69..ae3925c5 (commits 0dd2d5ad, ae3925c5)	ACCEPT. One should-fix (D1, landing condition) and 3 nits remain	combined file
ae3925c5..0e7e412a	ACCEPT. D1, D2 and D4 applied; D3 left for the pending §5 proposal	5f2058bf4a351d573fb22477e19a6d6b28d90684226a9c9a4b52f867a3de29c8 (both sections)

Landing condition (SF5/D1). #1083 lands only by a merge that keeps 8de75f7c and every later commit through the landed head reachable from main. A squash or rebase does not do this. The Q4 note's statement "§1 as frozen at 8de75f7c" and its revision record depend on those commits.

Delta check at ae3925c and confirmation at 0e7e412

## Delta check at `ae3925c5`

Object: `git diff 22ad8d69 ae3925c5` across both files. The range is two single-parent commits:
`0dd2d5ad` (parent `22ad8d69`) and `ae3925c5` (parent `0dd2d5ad`). The head is
`ae3925c55e7148e413432a1b43d01af5d5e9d9db`. The check was read-only and done by hand. I also
confirmed that #1083 c.6718607630 exists: it was posted from `amaybaum` at 2026-11-08T07:12:43Z and
carries this review.

### Verdict: ACCEPT

Remaining items: 0 blocking, 1 should-fix and 3 nits. The should-fix is a landing condition on the
PR, not a change to either file.

### Status of each finding

| Finding | Status |
|---|---|
| SF1 | Resolved. Q4-5 and Q4-6 now read "as Q4-2; (A-lin), or (A-aff) at μ_{A₁} = 1". |
| SF2 | Resolved. (ii) is stated "for O1–O4. … O5 is not examined". The outcome heading names amendment A5. The (iii) basis now reads "given H-b, on the lift, for a uniform offset v ≠ 0". |
| SF3 | Resolved, with nit D2 below. The controls line reports only what I verified: (C-a) as G_v = G_0 at v = 0, with the T_0 caveat under (A-aff); (C-b) from N0/N1; (C-c) from N3 and Q4-4; and the countercontrols N4a and N4b. |
| SF4(a) | Resolved. The R10 text is correct: under (A-lin) the loss is independent of resonance, and under (A-aff) it occurs only at μ_{A₁} = 1. |
| SF4(b) | Resolved, with nit D3 below. The D3 row and §6 Q4 now say "among constant linear unitaries ((A-lin))". |
| SF5, C2/C3 half | Resolved; see below. |
| SF5, ancestry half | **Open: should-fix D1.** |
| SF6 | Resolved. §8.2 states that §1 and §§3–4 entered together at `8de75f7c`, and that the precedence is the author's statement. |
| Nits 1–5 and 7 | Resolved. SM:318–322; the §5 owner quote "Retain linearity as a posit" is verbatim against c.6624263637; 𝒦 at :354; N6(a) now reads "Besides the A₁ phase (N1), …"; the Revision paragraph is complete; core Q1 now reads "posted from the owner's account (amaybaum)". |
| Nit 6 | Correctly left. §8.6 records that it was not applied and why. |

**§1.** Extracted from `8de75f7c` and from `ae3925c5` and diffed: byte-identical.

**Nothing strengthened; no new errors; no narration.**
- Each changed claim is the same as before or narrower.
- I scanned the added lines for history phrases ("now", "no longer", "formerly", "corrected form",
  "previously") and found none.
- The Revision paragraph and §8.6 are review-response records, which fits a WIP verification record.
- No manuscript file changed.

**C2 and C3 (core:73–74, :220, :540).**
- C2 now rests on SM:240 alone, with #1050's exhaustive review marked as provenance. This is
  accurate: SM:240 proves both the absence of a fixed point and the absence of any time-independent
  relabelling, and it states the remaining properties.
- C3 is narrowed to "a uniform offset on a ring of N sites with A − 2I = Δ (κ = 2), every odd prime
  q ∣ N" and cites B0.3(e). Re-checked by hand:
  - For q ∣ N the second-difference kernel is exactly {a + b·i}. It is periodic because bN ≡ 0, and
    it is two-dimensional because a second-order recurrence is fixed by x₀ and x₁.
  - Δ is symmetric, so Im Δ = (Ker Δ)^⊥.
  - 𝟙·𝟙 = N ≡ 0 and 𝟙·(i)_i = N(N−1)/2 ≡ 0, the latter for odd q. So Δv = −b𝟙 is solvable, and
    (v, v) is fixed.
- The wording is not circular. B0.3(e) now "is the support for C3" instead of "reproduces C3", and
  R2 says that C3 is its case B0.3(e) and that only the unauthorized §7 check would test it
  independently.
- What is left is nit D4 below.

**The D3 row and §6 Q4 narrowing.** I re-checked the statement "the stabilizer of the affine rule
among constant linear unitaries … is U(3)×U(2)":
- On the M-route with generic block-scalar M (Substratum:163), the commutant is G_0 and gv = v.
  This gives G_v = U(3)×U(2)×{1}.
- On SM's Σ-route the linear symmetries of the rule are U(5)×1, and their intersection with Stab(Σ)
  is G_v.
- So the narrowed statement is correct on both routes once Stab(Σ) is included. See nit D3.

**Decision on D3(iii) (core:147–149): it suffices, with an optional nit.**
- D3(ii) (core:145–146) fixes the domain: "a unitary g acting identically at every site". A unitary
  is linear. So (iii)'s "the stabilizer of the affine rule is U(3)×U(2)" is the (A-lin) statement
  and is true as written.
- The (A-aff) branch, under which the statement fails off resonance, is outside that domain.
- The one residual ambiguity is (iii)'s bold lead, "The affine stabilizer loses the singlet U(1)".
  It can be read as the stabilizer under affine maps, the reading that review B1 separated out.
- Q4 §5 already proposes the remaining D3(iii) refinement, naming the route and adding the
  resonance condition, for the owner.
- Leaving it unchanged is acceptable. An optional minimal edit is to read the lead as "The affine
  rule's unitary stabilizer loses the singlet U(1)".

### New items

**D1, should-fix (landing condition; carried over from SF5).**
- The ancestry half of SF5 is not reflected in either file. §8.6's SF5 entry describes only the
  C2/C3 half, while its lead says the findings were "Applied".
- Two things are needed:
  - Record on #1083 that the PR lands by a merge that keeps `8de75f7c` .. `ae3925c5` reachable,
    not by squash or rebase. Alternatively, quote §1's hash at `8de75f7c` in §8.2.
  - Add "the landing condition is recorded on the PR" to §8.6's SF5 line, or state there that it
    is left to the PR.
- If the condition is already on the PR, only the §8.6 wording remains.

**D2, nit (Q4 §4 controls line).**
- "N4b, an offset in T₁, which loses SU(3)" omits N4b's own condition: under (A-lin), or under
  (A-aff) only if μ_{T₁} = 1.
- *Fix:* add "(under (A-lin); under (A-aff) only if μ_{T₁} = 1)".

**D3, nit (core D3 row).**
- The row cites both Substratum:163 and SM:344, but does not say that on SM's Σ-route the result
  needs the intersection with Stab(Σ). Without it, the constant linear unitaries give U(5)×1.
- This is the same "name the route" point that Q4 §5 already proposes for D3(iii). The fix can wait
  for that proposal, or add "(on the Σ-route, within Stab(Σ))".

**D4, nit (core C3).**
- C3 is now a derivation made in this document ([D] B0.3(e)). It sits in a table headed "Results to
  be preserved, which serve as controls", so it is an internal consistency check, not an
  independent preserved result.
- R2 already says that only §7 would test it independently. A two-word label in the C3 cell, such
  as "internal check", would make that visible in the controls table too.

Residual, no change needed: §8.3 still says the [O] label is "kept only on that quotation", although
§5 now also carries a sourced [O] quote. Like §8.4 (nit 6), §8.3 is a response record of an earlier
revision.

***

## Confirmation at `0e7e412a`

Object: `git diff ae3925c5 0e7e412a`. The commit is `0e7e412a42401616d5c2f1829a20b78becb828a4`, and
its single parent is `ae3925c5`. The diff touches two files: 5 lines added, 3 removed.

**Verdict: ACCEPT.**

- **D1 is applied.** §8.6's SF5 line now says that the commit-ancestry part is a landing condition
  recorded on #1083: the PR lands by a merge that keeps `8de75f7c` and its successors reachable.
  The wording is correct.
  - The line says the condition *is recorded* on #1083. I could not check that, because this pass
    was read-only with no fetch, and the coordinator reports that the PR body is being updated now.
  - It must be true before landing: the PR body has to carry the condition.
- **D2 is applied.** The §4 countercontrol now reads "N4b, an offset in T₁, which loses SU(3)
  (under (A-lin); under (A-aff) only if μ_{T₁} = 1)". This matches N4b and Q4-7.
- **D4 is applied.** The support cell of core C3 now reads "[D] B0.3(e); an internal check, not an
  independent control". This is accurate.
- **Nothing else changed.** The diff holds only these three edits, and no other line moved.
- **§1 is unchanged.** I extracted it at `8de75f7c` and at `0e7e412a` and diffed the two: they are
  byte-identical.
- **D3 is left for the pending §5 proposal,** together with the optional D3(iii) lead wording. That
  is acceptable: both are optional refinements, and neither makes a statement false.

---

Generated by Claude Code

[amaybaum]

Merged as 9e0560d8. This was a normal merge with parents 924d421b and 0e7e412a. All commits from 8de75f7c through 0e7e412a are reachable from main. Only the two A5 files were added. Post-merge main CI run 42459408153 is green. The integration record is on #1049.

---

Generated by Claude Code