[amaybaum]

Composition and observer-access research record. All results are internally reviewed research. None of them is a native gate, native storage, a native reset, an observer-access rule, a reusable instrument or full OI-to-QM equivalence.

• Head: 2af1d5d5b5fff4375e9fe9460f892c680ae99f0c.
• Everything is under verification/programmes/oi-qm/reconstruction/composition-access-20261024/.
• Package index and replay;`` handoff.```

Merge: not authorized.

Results

Conditional composition theorem (top-level REPORT.md)

• Suppose each stage's complete-record kernel intertwines a fixed positive decoder with the reference instrument on an admissible cone, and initialization and physical/access closure hold. Then every finite adaptive protocol over that menu has the reference full record law and outgoing decoded states.
• This is a written induction. It is a sufficient contract, not a characterization.
• Exact toy controls:
  • 12 local cases;
  • 324 policy and initial-state cases;
  • joint-law, heralded-seed, calibration, failure and access countercontrols.
• Accepting review at a7f22c38.

Applications to native constructions (native-x-v2/, tagged-post-y-v3/, native-phase-v4/, native-coupling-v5/). Each has an accepting review.

• Native-X candidate.
  • It gives a conditional one-stage success-branch instance on the supplied preparation cone, and the terminal failure law.
  • Offering the unchanged carrier after its terminal read leaves the decoder cone (determinant −1/4).
• Tagged post-Y.
  • A fixed tagged decoder accommodates the known-history tag.
  • Unchanged post-Y samples leave the positive cone after X, R or Z reads.
• Native phase. This is a finite-depth operational comparison.
  • In local boundary controls (reachability not claimed), an omitted native neighbour's parity changes the next raw record.
  • A matching next-record law can still miss a joint record/carrier defect (TV 1/2).
• Coupled pool.
  • In the scaled two-source control, native propagation correlates the pool, and exposing the whole current q9 port gives nonpositive accepted decoders.
  • Native A's third source gives a conditional code/metadata law in which the whole current q9 record is independent of the offered carrier.

Logical acceptance circuit (composition_flag_v6/)````

• The circuit. An explicit 89-tick NOT/CNOT/Toffoli circuit takes the 21 bits of the actual phase as logical input wires.
  • It computes the acceptance predicate, copies the flag, and restores all phase and temporary bits.
  • It uses 31 core bits, of which 8 are clean and 2 initially fair.
  • An optional record-bank dilation uses 301 bits and 359 ticks.
  • These figures are upper bounds on supplied logical schedules, not native runtimes or minima.
• The joint law. Conditional on Native A's source contract, its complete 90-sample pointer prefix and outgoing carrier have the intended joint law, with every failure and zero cell retained.
• Supplied, not shown: native gate compilation, operand storage, clock and workspace isolation.
• Accepting review at 1cb224db.

Delayed old-port history (delayed_history_v7/)`````

• The statement. For every L < H ≤ 2L, the complete public record is jointly distributed with the outgoing carrier as the v6 law tensored with an independent uniform tail. It is conditional on:
  • the hypothetical serial alignment of v6 with native ticks;
  • the Native A source contract;
  • the supplied history_v11 bank and G priors;
  • v6's supplied isolated logical storage, and its access contract with the eight old ports added.
• At actual L, this is a written conditional corollary of exact integer geometry and the inherited results. The exhaustive cell-by-cell check is a scaled control (n = 3), and it does not establish the actual-L statement.
• No native storage: a natively stored carrier can see tail-bank triples from L+77.
• Process. The plan d72ea524 was committed before computation. Reviews blocked 114ad502, 74cdd24e and, on re-check (x4), 98cf4647, then accepted 65650298 with nits.

Passive native readout (native_predicate_v8/)``````

• The result. In the scaled control (n = 6, sixteen dirty families), no read set in class C(4, 16) gives the complete v6 law conditional on C0. It is NEGATIVE_FOR_CONSTRUCTION, for that class only.
• Not excluded:
  • fresh pads, prepared sources, gates, larger regions or later reads;
  • laws holding only after averaging over backgrounds;
  • actual L.
• Process. The plan 631bec0c was committed before computation. Reviews accepted with nits at 375a29e4, then accepted at fc452ea8.

Landing preparation

4d46eeeb, c2f22f9d and 2af1d5d5 change only README.md, CURRENT-HANDOFF.md and the top-level SHA256SUMS line for README:

• README becomes the package index, with each package's plan and review coverage, root-relative replay commands and the provenance of the inputs;
• the handoff no longer restates WIP: native nonlinear processing and second-front source audit #910's or States and dynamics: exactness, sourcing and finite-approximation audit (research record) #911's results, and drops dated and process sentences;
• the native-coupling-v5 provenance pin is corrected forward (below);
• c2f22f9d applies the scope wording from the independent claim check (below) and adds three disclosures to README:
  • tagged-post-y-v3's plan and checker share commit 75619cb7;
  • the top-level review is review/;
  • v5's cited F9 plan commit 757e766b changes no file.

No checker, result, plan, report, review or package manifest changes.

Provenance pin.

• native-coupling-v5/PROVENANCE.json pairs shared_interface_commit 1324d9a8 with the digest 1a0b34eb…. That digest is of shared_v12/INTERFACE-CONTRACT.md at 05397c79; the file at 1324d9a8 hashes 95cd4d20….
• The package uses the two-slice identity that 05397c79 adds, so the contract it rests on is the file at 05397c79, which is unchanged on main. composition_flag_v6/PROVENANCE.json records the same.
• The provenance file is pinned by its review and stays unedited. README records the correction.

What was checked

• Independent claim check.
  • A separate-agent internal check of the eight complete-record composition claims at 4d46eeeb found that all hold within their declared scopes, and every cited number replays byte for byte.
  • Its six wording corrections to README and the handoff are in c2f22f9d.
  • Its re-check of c2f22f9d accepted, with one provenance sentence corrected in 2af1d5d5: 1b729355 appends the third-source fork to the existing native_x_v12/PLAN.md.
• Merge. git merge-tree with main 8b5e8681 is clean. The landing adds 298 files, all in the research directory, and modifies or deletes none. No path is shared with WIP: native nonlinear processing and second-front source audit #910's or States and dynamics: exactness, sourcing and finite-approximation audit (research record) #911's directories.
• Replay.
  • All 14 checker/result pairs (8 author, 6 review) are byte-identical under python3 -I -B, normally and with -O, by the README commands from the repository root.
  • The v8 mutant campaign reproduces mutants/mutants.out.
  • Reviewer scripts reproduce their retained outputs: the v7 round-2 mutant summaries, v7 round-1 review_checks.py, and v8 round-1 rederive.py and coarse_probe.py. The one difference: v8 rederive_allv.py prints two histogram lines per seed that its retained output omits.
  • All 16 SHA256SUMS files pass (281 files).
• Pins.
  • Of 66 contextual pins (review input hashes and blob ids, provenance snapshots), 65 verify at the commits they name. The exception is the pair above.
  • All 20 commit/path/sha256 triples verify.
  • Every file a review pins is unchanged at the head.
  • The nine WIP: native nonlinear processing and second-front source audit #910 commits cited are on main, and the WIP: native nonlinear processing and second-front source audit #910 files read for results are unchanged there.
• Gate. On the scratch merge, 20 of 21 steps pass; lean-axioms needs lake, and this PR touches no Lean. Legacy-records: 303 intact. V3: self-test, 140 corpus vectors, 52 receipts hold.
• Reachability. After a merge commit, every commit the directory cites is reachable from main except 2d555f17.
  • 2d555f17 is on closed Research (WIP): the missing component for F2-reg, and measurement as an added condition #885's branch research/missing-component-20261020, cited as provenance for the programme acceptance rule.
  • main already cites 0965d1a0, this branch's start, in native-y-continuation-20261024/shared_v12/INTERFACE-CONTRACT.md. It becomes reachable with this landing.
• Hygiene. No secrets, e-mail addresses, IP addresses, local paths or binaries. The one recorded redaction removes an absolute scratch prefix from four reviewer err.txt files.

Landing instructions

• Land with an ordinary merge commit, main first and the head second, so the cited commit ids stay reachable. Do not squash or rebase.
• Keep Research (WIP): the missing component for F2-reg, and measurement as an added condition #885's branch research/missing-component-20261020.

Related: #910 (native construction, merged), #911 (states and dynamics, merged), Issue #913 (wider gates).

🤖 Generated with Claude Code

https://claude.ai/code/session_013T6pfKuv5mi3MXtKjH2tZb

[amaybaum]

Closing without merging (D-69: not landed now; PR-AUDIT-1, #1088). The branch and its commits are kept.

Owner decision (amaybaum, Claude Code session chat, 2026-11-04):

Approved, with one administrative condition concerning O-REUSE. I accept the final independent confirmation and approve retiring both #935 and #912 without merging. Neither PR's scientific results are adopted by this decision.

D-69 — #912: Do not land now. Close without merging.

A future integration of #912 requires a separate owner decision addressing the strict up-to-date checks, exact-head review, and preservation of cited commits. This approval grants no bypass.

This is an administrative retirement, not a scientific adoption. #912's conditional findings remain branch-only evidence. No native gate, storage, reset, observer-access rule, reusable instrument or OI–QM equivalence is established, and no scientific status changes.

• Accountable homes:
  • OI reconstruction: remaining foundations, generalization and A2 gates #913 (handoff comment): the eight packages with their accepting reviews; the claim-check chain 4d46eeeb→c2f22f9d→2af1d5d5; the v5 pin correction at 4d46eeeb; the rederive_allv.py replay note; the open gates. It also holds O-REUSE (tier Foundational). The transfer from OI ↔ QM equivalence: proof gate (dependency index; carries #885) #914 is complete: OI reconstruction: remaining foundations, generalization and A2 gates #913's planning block carries the row, and OI ↔ QM equivalence: proof gate (dependency index; carries #885) #914 recorded the supersession (comment).
  • NP-AVG-1: does the native acceptance predicate's negative depend on the pointwise background convention? #929: the averaged-background branch of v8, and the NP-AVG background-convention part of O-REUSE.
  • OI ↔ QM equivalence: proof gate (dependency index; carries #885) #914 (routing update): Research (WIP): the missing component for F2-reg, and measurement as an added condition #885 obligations 5 and 6 (row O-INS; their work continues in OI reconstruction: remaining foundations, generalization and A2 gates #913's gate rows), the re-homing, the C-14 pin (grade PD unchanged), and decisions D-66 (open) and D-69 to D-71.
• Provenance kept:
  • branch draft/oi-composition-access-20261024 at 2af1d5d5b5fff4375e9fe9460f892c680ae99f0c, with all 298 files and 40 commits;
  • 0965d1a0, which main cites at native-y-continuation-20261024/shared_v12/INTERFACE-CONTRACT.md:12;
  • Research (WIP): the missing component for F2-reg, and measurement as an added condition #885's branch research/missing-component-20261020, holding 2d555f17, which main cites.
• Why nothing is landed: no claim on main rests on these records as support. main names Composition and access: finite-depth composition contract, logical acceptance circuit and complete-record laws (research record) #912 only as a pointer: "results not established here".
• Why the branch must stay:
  • it is WIP: NP-AVG-1 Stage 0, the masking geometry of native records (planning, #929) #935's base, and reopening WIP: NP-AVG-1 Stage 0, the masking geometry of native records (planning, #929) #935 needs it. While any PR still targets it, deleting it from the Branches page or the API makes GitHub close that PR (documented behaviour);
  • it holds fc452ea8, which WIP: NP-AVG-1 Stage 0, the masking geometry of native records (planning, #929) #935's frozen plan and NP-AVG-1: does the native acceptance predicate's negative depend on the pointwise background convention? #929 require.
• If the work lands later: reopen this PR and land it by ordinary merge only, main first and this head second. Do not squash or rebase.
  • main's ruleset guardrail requires three status checks under a strict up-to-date policy, and this PR is behind main.
  • Landing at 2af1d5d5 therefore needs either an owner-authorized bypass of the up-to-date rule, or an update that moves the reviewed head, followed by a fresh review and check.
  • GitHub may offer Update branch. Using it is the second path, and only with that decision.
• Scientific status is unchanged by this close.

---

Generated by Claude Code