[amaybaum]

This paper screens one register for #1032's owner-authorized classical O5 design. The source check stops at missing physical register/read access. The existing PAD or NP obstruction also covers every retained preparation law for this one geometry, so the paper does not supply the requested unresolved retention tuple or a second-hold result.

Paper-only; draft and unmerged. Independent exact-head ACCEPT as a stopped screen only. Physical retention remains open.

• Current head: 3ec25f1fbe18e822924f6e834f8fa072a8f4244c.
• Base: main f49b768709d262a05471895cec703b31aff19b57.
• First fixed-candidate checkpoint: cc2d21b36640c421d28dcca76840e97ba26e6675; the completed screen e636d1d051bf775d26ab86fd2530c23da095d90b and current citation correction are successive single-parent commits.
• Owner scope: #1032 c.6692541241. One candidate only; a physical geometry/access source must precede retention analysis, otherwise stop.
• Candidate: SMR binary ring at N=8, complete incoming stencil A={7,1,2}, same preregistered nu_r laws. A structural stencil is a motivation to screen, not a physically readable apparatus.
• Scope: one new paper, verification/programmes/oi-qm/reconstruction/o-ret-routes-20261103/phase-b-o5-20261106/DESIGN.md; no other source, frozen input or historical research(#1032 O-RET): Phase B native-retention design (design only; WIP / UNVERIFIED) #1115 branch changes.
• Counting: corrected criterion restated and provisionally adopted for this classical design. Ambiguous D1/D2 and write-tick input cases remain undecided. No storage row or counted pass is claimed.
• Other obligations: cross-read coherence remains separately blocked in O-RET: physical one-step retention and carrier-sensitive native readout (F2 planning) #1032; OI-QM-SEL: finite-dimensional operational selection and native distinguishability #1143 owns its distinguishing test. P0: physical observable carrier and evolution selection #866 keeps B3's outcome-conditioned physical exchange and public access; no new OI-QM-EXT-1: establish conditional equivalence with explicit added conditions, then minimize and source them #1169 assessment is authorized.
• Checks: document duplicate/whitespace/inline-code and all six relative source-path checks passed; remote paper read back exactly; GitHub compare confirms only the new paper differs from the base. Ordinary exact-head PR CI 42275195143 passed at 3ec25f1fbe18e822924f6e834f8fa072a8f4244c (25 successful jobs, 1 skipped).
• Independent review: full review and exact-head delta ACCEPT at 3ec25f1f. The sole required source-role/reachability correction is closed. Acceptance is limited to the stopped-screen record; no unresolved candidate or physical-retention demonstration is supplied.
• Next gate: owner disposition of this stopped record and a separately bounded scope if further candidate work is wanted. No replacement candidate is authorized by this PR. No freeze, implementation, scientific computation, enumeration, simulation, Lean work, replay, dispatch, new physical premise adoption, merge or claim promotion.

Accountable Issue #1032 stays open.

[amaybaum]

Completed one-candidate source screen, independent review requested at e636d1d051bf775d26ab86fd2530c23da095d90b. Parent is the fixed-candidate checkpoint cc2d21b36640c421d28dcca76840e97ba26e6675. The source gate stops at missing physical joint read access to y_{1,2,7}. The existing PAD/NP applicability check independently prevents claiming an unresolved O5 tuple: pi0 is padded; for all ready laws the complete incoming stencil is zero, so NP's uniform-G0 exception is not met. No second-hold trajectory or counted pass is claimed, and no replacement candidate is selected. The paper retains the provisional counting policy and the unresolved D1/D2 write-input cases. Static document/source-path checks passed; exact-head ordinary CI 42273931949 is pending. Please review this only as a stopped screen, including whether any language overstates the owner's intended unresolved O5 design. Freeze/run/merge remain outside scope.

[amaybaum]

Independent model-authored paper review — ACCEPT WITH CHANGES, as a stopped screen only.

Reviewed exact head e636d1d051bf775d26ab86fd2530c23da095d90b, against main/base f49b768709d262a05471895cec703b31aff19b57. I am a separate Codex reviewer who did not author this paper. I independently read the pushed DESIGN.md, the owner's exact ruling, and the applicable AGENTS.md rules, including §§A.19, A.40 and A.42–A.46. This is paper/static verification, not an external human review, scientific run, or CI attestation.

The substantive outcome is correct: this geometry has no sourced physical joint read, and every retained preparation law is already obstructed. It supplies neither the requested unresolved retention tuple nor a second-hold result. Accepting its candid stopped-screen record must not be represented as accepting a physically sourced O5 design or completing the owner's unresolved-candidate objective.

1. The fixed mathematics and preparation family check out. SM §3.1's switching rule reduces over the binary alphabet to the stated OR/XOR rule. The full current-slice parent set of the carrier update is indeed {7,1,2}. The prior checkpoint cc2d21b36640c421d28dcca76840e97ba26e6675 already fixes this same geometry, all four carrier preparations and the unchanged nu_r family, while recording the source/escape screen as pending. The reviewed head is its single-parent successor, and the PR changes only the one new paper. This verifies the published checkpoint ordering; it is not evidence about unrecorded private analysis. The incoming stencil is a structural reason to screen the shape, not a sourced observer or apparatus.

2. I independently checked all six obstruction entries. For this locus the structural outside-parent set is Q={y_0,y_3,y_4,y_6}; it cannot be pruned to the prepared trajectory. Under pi0, the independent uniform previous register slice pads the write, so PAD gives a blind readout and worst-case error at least 1/2. Under every ready law, all of y_7,y_1,y_2 initially vanish, hence G_0=0 and y_0(t0+1)=x_0. Since y_0 is in structural Q and both values of x_0 remain independent preparations, J-i fails at the first hold tick. This directly confirms NP's obstruction, without any decoder choice or second-hold evolution. NP applies through its own hypotheses even though this non-arc geometry is outside Theorem N's original family. AE's global fixed-coefficient hypothesis fails because y_0 OR y_1 = y_1 XOR ((1 XOR y_1) y_0) at site 7; readiness on one initial support does not repair the global hypothesis. LS fails through the non-arc shape; U fails because sites 1 and 7 structurally read y_0; SS-2 requires width 1. Thus escaping those four hypotheses does not escape PAD/NP. No wider O5 no-go follows.

3. The physical source stop is properly scoped. I inspected the stated main anchors, performed repository-wide source searches, and also checked the candidate-adjacent native apparatus/operand records. SM supplies the update, Main supplies observer conditions and representation/existence statements, and the ready-source and native-pointer records retain their qualifications. None of the inspected material supplies the particular embedded process that jointly reads y_{1,2,7} with apparatus state, resources and back-action accounted for. Generic representation or coupling results do not identify that process. The paper correctly states a source-search finding, not an impossibility theorem; E-SDC, declared H-cur and readiness remain unsourced/unadopted. The owner's explicit source-failure stopping rule permits this stop. It does not make this already-obstructed choice an unresolved candidate.

4. The unperformed second-hold step is not concealed. The stated T=2 convention is exact: hold ticks t0+1,t0+2, read window t0+1,t0+2,t0+3, tolerance 1/20. A value at the second hold tick alone would be insufficient. Given the failed source gate, the paper appropriately supplies no new trajectory, decoder, error table or storage verdict. This review neither requests nor authorizes such an analysis to compensate for missing physics.

5. The provisional policy is faithfully restated, with important limits preserved. I compared §6 with historical DPB §5 at 4b4cacc93884039afadd0069373b0ac448761a6b and D1/D2 at 43370576 §6.2. D1's minimax/lexicographic rule and zero off write support, D2's full-configuration relation and conflict fallback, structural Q, hold-input support union, and the corrected input exclusion supp Q_t0 union supp Q_(t0+1) agree. The IP implication is independently checkable from these definitions: if every hold input is one used in constructing D2's relation, every supported register state remains in a class anchored at the write support, so the decoded value is preserved. IP-prime's later class-merging property is not a substitute for this test. The D2 write-tick-only case and the D1 case whose entire hold-input support already occurred at the write tick remain undecided, as the owner required. Merely saying that D1 reads no hold-tick data does not remove that caveat. The free-map alternatives are logical calibration alternatives, not sourced native dynamics; the image/nonconstancy qualification is retained, and AE-D2 is not improperly applied to this non-AE geometry. There is no claimed counted pass here to certify. This review does not upgrade the provisional rule into a universal theorem that every future flagged storage row can fail the complete thresholded classifier. Any future counted claim still needs its actual allowed failure alternative or a sufficient argument for that actual criterion, including the write-input caveat.

6. The other source boundaries and obligations are preserved. All six relative source-map paths resolve to files at the stated main base; Theorem N's companion review and the frozen Phase A window/threshold were checked. Historical decoder material is restated as working policy rather than used as an off-main proof of retention. The OI-QM-SEL: finite-dimensional operational selection and native distinguishability #1143 and P0: physical observable carrier and evolution selection #866 comments are planning provenance for the separate obligations, not scientific premises. Cross-read coherence remains separately blocked on a sourced coherent carrier and preserving read; an assessed two-read match alone does not identify that mechanism. B3's physical outcome-conditioned exchange and subsequent public access stay with P0: physical observable carrier and evolution selection #866. No physical-retention conclusion is inferred from the ideal B3 model.

Required narrow correction: DESIGN.md line 234 calls the first checkpoint “Provenance” although lines 7–8 use it to support the ordering claim. AGENTS.md §A.42 item 8 explicitly says that a claim depending on commit order or ancestry uses those commits. Label this reference as support for that specific ordering claim, currently in PR ancestry, and make explicit that any separately authorized landing retaining the claim must make that checkpoint reachable from main. Merely preserving its branch is insufficient; alternatively, withdraw the ordering claim. The existing reachability warning recognizes the issue, but its source-role label and destination need to agree with the rule. This correction changes no scientific conclusion and requires no replacement candidate.

After that correction, obtain an exact-head delta check. This review authorizes no freeze, implementation, enumeration, simulation, Lean/scientific execution, replay, dispatch, premise adoption, merge, publication or scientific promotion. I changed no files or refs and ran no scientific computation. Ordinary PR CI remains separate verification and was not used as evidence for the mathematical or physical findings above.

[amaybaum]

Independent model-authored delta review — ACCEPT, as a stopped paper screen only.

Reviewed exact head 3ec25f1fbe18e822924f6e834f8fa072a8f4244c, the direct single-parent child of e636d1d051bf775d26ab86fd2530c23da095d90b, against my full independent review. I am the separate Codex reviewer who did not author this paper.

I inspected the actual GitHub compare, commit parents, current PR head and exact pushed file. The delta changes only two passages in DESIGN.md. An exact text comparison confirms that all other file content is unchanged.

• The sole required correction is closed. The checkpoint row at line 235 now correctly identifies cc2d21b36640c421d28dcca76840e97ba26e6675 as support for the ordering claim, currently off main in this PR's ancestry. It explicitly requires reachability from main no later than any separately authorized landing retaining the claim, and correctly says that preserving a branch alone is insufficient. The conditional statement about a normal merge preserving the current ancestry is accurate and grants no merge authority.
• The header change is appropriate. It retains BLOCKED status and replaces the expiring “review pending” statement with a stable pointer to the guiding PR's independent model-review record. The prior review is present there.
• No scientific or scope delta exists. Geometry, preparation laws, obstruction checks, window, threshold, provisional counting policy and its write-input caveats, source stop, and separate coherence/P0: physical observable carrier and evolution selection #866 obligations are unchanged. The findings and limits of the full review therefore carry forward.

There are no remaining required changes from this review. Acceptance concerns the candid stopped-screen record: physical joint read access remains unsourced; PAD or NP covers every retained law; no unresolved candidate, second-hold result, storage row or counted pass is supplied. It does not complete the owner's unresolved-design objective or authorize a replacement candidate.

No files or refs were changed by this reviewer, and no scientific computation or workflow was run. Ordinary CI is checked separately. This verdict provides no freeze, run, implementation, premise-adoption, merge, publication or scientific-promotion authority.

[amaybaum]

Owner normal-merge approval and ancestry requirement: #866 c.6718471099. Previous #1185 post-merge main CI 42458124176 is green. Current main 6a987a0ba9898f2e0a9115dd5854a04725473078; unchanged reviewed head 3ec25f1fbe18e822924f6e834f8fa072a8f4244c, full/delta independent ACCEPT and exact-head CI 42275195143 green, raw GitHub data mergeable. API confirms merge base f49b7687; the conflict-free integration tree is 20cc54ea0ab9fe681060749f694565ac2323fd64, with a byte-identical reviewed one-file/242-line landing delta. The exact chain cc2d21b3 → e636d1d0 → 3ec25f1f was verified; an ordinary two-parent merge preserves all three and makes the ordering-support checkpoint reachable from main. No squash, rebase or cherry-pick. The stopped O5 scope and unresolved policy/coherence/source obligations remain. Wait for green post-merge main CI before #1186; #1032 stays open.

[amaybaum]

Normal merge 924d421b0e8b40d276196eb5aab7e5bec6865970 has the exact predicted tree and both expected parents. The host confirms supporting checkpoint cc2d21b36640c421d28dcca76840e97ba26e6675 is an ancestor of main, preserving all three reviewed commits. Post-merge main CI 42458937435 is green: 25 successful jobs and one conditional skip. The stopped O5 scope is unchanged; #1032 stays open. Concurrently, Claude's #1083 added two separate A5 papers as 9e0560d8; the next integration preserves them.