[ToikY]

Change and mathematical scope

This independently verifies the first asynchronous helper split of PR #5786, rather than claiming its construction or conditional exponent. It binds the actual temporal contacts to the exact PR #5774 word, emits the first split in an independent label namespace, and checks every arbitrary-input endpoint and frame/owned-lifetime condition.

The result is c=e696^T, d=e696+e697+e699+e709+e733+2e133, with unit pairing. The literal witness checks 791 formal input axes, 625,681 endpoint positions, 25 owned lifetimes and peak 12 live owned fields. The difference buffer is explicitly z-x_h before its charged CLEAR. The original 789-axis source endpoint is checked separately on all 622,521 positions.

The owned contact at time2=27501 subdivides query13750, MOVE770:591->597, through frame1031. Actual dimensions are 8 > 5 > 1, hence paid ranks 7 -> (3,4) and access-query delta H3+H4-H7. This is only the access query, not a full invoice. Helper target ranks are 5 and 4. Native maps/absolute phases, complete scalar/buffer/route pricing, later steps, physical caller, compiler and full admission remain open. The five framework assumptions are retained; no new kappa or frontier change is claimed.

Task and overlap

Main: de5ec2e7a6074068452f45b1a191b5c72832df4a, priority 8, refreshed before submission at 20:42 UTC. Exact predecessors: #5786 6b936acd68a7e0f9062037900f0fe9bc4b6fdd25; #5774 7d6ec0cdef31d0c37b7d2563e709979caa0d5eae.

This answers the released first-step review request. Our claim changes the formal-interface-lemma card to this finite exact temporal-interface audit; no Lean formalization or framework-assumption closure is claimed. It differs from the active different-output-owner construction and the author's full replay. The final open-PR refresh covered 352 records; Discussion coverage is partial and unpublished work remains unknown.

The diff is exactly one 12-file package, research/p104-async-first-hub-audit. It vendors only the frozen word/plan data and retained NOTICE; neither checker imports predecessor implementations or the other checker. Source origin commits, paths, raw SHA256/Git blob identities and canonical word SHA256 are recorded. Self-contained data avoids reliance on cached unmerged commits.

Validation

Python 3.12.14, standard library:

python3.12 -B scripts/research_package.py --package research/p104-async-first-hub-audit

Registered normal/optimized primary and opposite-direction oracle pass. The oracle reconstructs collection by reverse covector and delivery by four independent forward perturbations, then checks source columns/pivot row for basis commutation. All typed receipts agree, with canonical JSON comparison preserving boolean/number distinctions.

Seven internal controls reject omitted collection, owned injection, causal order, wrong source binding, unpaid access, new buffer CLEAR and a swap sign; the last two fail at endpoint identities. Separate source/receipt corruption and boolean-number alias controls also reject; replacing canonical comparison with ordinary Python equality admits those aliases. These deliberately injected faults are controls, not defects alleged in predecessor code.

Metadata/pins, committed bytes/modes, declared current-main history and official export integrity pass. A separate read-only, Git-free -I -O -W error -B replay with empty PATH passes both checkers in about 1.42 seconds, with every exported byte/mode unchanged. Official export SHA256: e582305b536fca3d18a6e4fd83acced78fa1b70dd34115ca8d5c90f2b02e7ddd. The declared history check covers the main-context pin; original data origins are authenticated by retained bytes, not claimed as publication ancestors.

Current selected baseline passes bindings/metadata only after exact input materialization. No selected native replay ran. Hosted checks have not yet been observed; existing main's shared assignment-test failure is addressed separately by #5761. No hosted scientific pass is claimed.

Publication head: c28110134390499b181ad62f5047514da4f3d5d1; tree 0911eb0070abfa7ff8213d3aec535b1e417cf027. This is an independent-label witness, not byte identity with the author's emitted word.

Credit

Ali Romman, with OpenAI Codex assistance, supplied the asynchronous mechanism, plan and retained-copy source. Retain flyingviper's boundary bases, Joel Pulikkan/GamingPuzzled's basis work and every native/controller/compiler/BIT predecessor in NOTICE. New audit/proof: ToikY with substantial OpenAI Codex assistance, using separate implementation, independent-oracle and read-only review contexts. Apache-2.0.

Open next questions

Check native maps and absolute phases for the owned subdivision, then bind all boundary/buffer/scalar/route costs to the first-step full invoice. Steps 2/3 must use their own exact input words and event indices. TASK.json records the structured follow-up. No merge or closure is requested. Maintainer edits disabled.

[ToikY]

Hosted CI update for exact head c28110134390499b181ad62f5047514da4f3d5d1: the planning job failed in the unchanged shared test_research_current_assignment.py suite: eight failures and two errors, with 22 platform skips in 511 tests. The scientific replay job was skipped. The revision-8 fixture repair is already proposed separately in PR #5761; this focused audit does not import that unrelated repair. The entire diff is confined to the new package. Registered normal/optimized primary and independent oracle, committed-byte/mode/history checks and read-only Git-free cold replay pass locally as recorded. No hosted scientific pass or independent admission is claimed.