Reduce helpers 644/660 from width 8 to 7; conditional κ > 0.00173

[RossRichardson]

Summary

Change the actual PR3133 scalar words and entrance spaces so helpers 644 and 660 have ordinary width 7 instead of 8, then realize the changed inventory with 740 replicas, 6,660 calls and 2,717,280 complete chronological owner roles. This changes the construction and its complete paid histogram; it is not a decimal-precision update or the earlier proposed 7→6 extraction.

The resulting exact complex root lies in [0.001730006549538841847005, 0.001730006549538841847006]. Under the documented inherited assumptions and the source-bound full release, the modern joint transfer gives κ = 0.001730006549538841847005 − 10^(-46) − 10^(-70) > 0.00173. The complete original strict release and separate exact source/native/owner/physical/prefix/outer binding both pass.

The independently runnable local word/native package is unchanged. The sibling release package adds a 21.8 KB compressed integer primal, a standard-library generator for the literal owner journal, an independent all-row audit, a complete conditional derivation and an explicit long source/native/physical/prefix/outer replay recipe. Canonical scope and status.

Scope and limits

The finite checks cover both arbitrary-input orientations, all 91 private buffers, exact native phases, source-window and physical-family-safe owners, every chronological pending prefix, the full eight-stratum histogram, and rigorous modern/bootstrapped outer arithmetic. Physical stock is 663,549; the actual prefix is 1,479 bits, the sufficient reserve is 1,512 and the paid reserve is 2,048.

The conditional implication retains the all-size dispatcher/compiler, uniform coefficient/stock and SCALAR/COPY/CLEAR/zero-allocation contracts, common-label/mutable-source physical realization, whole-call hosting, internal recursive rounding, tape costs, retained bootstrap-time constant, exact recovery, PR559 recurrence/correction, analytic transfer and BIT H1–H5 hypotheses. No unconditional theorem, formal-kernel proof, global packing optimum or selected-frontier change is claimed.

The frozen historical source templates are preserved. Their old 3000/3080-digit prose and 15-control summary do not replace the actual precision-24 SPEC/LEDGER or 25-control ASYNC-AUDIT. The new long wrapper's preparation stage was executed; its full tail was not. The original complete release run separately passed and is bound by exact source and validation manifests. Fresh retained bootstrap is supported by the recipe but was not rerun in this session; explicitly reused retained source/receipt/cache bytes are authenticated.

Test plan

• Original local package: full registered replay, optimized Git-free committed export replay, exact byte/source-history gates, and automatic CI39111312157 all pass on ac92a027c78b0c9bb0ada739b26959afda489d0e.
• Sibling: registered fresh owner/moment replay passes. The committed source-history and byte gates pass at c006ddb034f5efcb0706a5d8207d38c367a99ace (145 declared files). Its Git-free export replay passes under -I -O in 63.83s with all 146 regular files, including the export receipt, unchanged. Exact archive SHA256: 8a021a5280b615ad353cd6d0ba489e815ad9995413d72af40584db3b2bf0a692.
• Three exact moment engines, all 63 modern outer inequalities and 17 controls, and all 47 bootstrap inequalities with four controls pass. Four aggregate corruptions and five separately scoped literal predicates are checked; the complete literal journal is independently audited.
• Portable long preparation passes exact historical exports/source rebuild, both fresh native chains, fresh owners and actual owner/lift/rebind preparation. The original full strict release passed in 1,138.41s; the independent binding authenticates actual 25 controller controls, exact source and complete histogram. Automatic CI39113623065 passes on publication head 3efbb5de06a49ffeea6684482f2d3fe48cdca3fb, including both the unchanged local native package and the new conditional-release package. No manual CI rerun is requested.

Sources and credit

Pinned main 20ecf2891d80b5acf18d980b2164fb44ea7faa90; dcharb78/PR3133 51eb9c4ee9e4722010d308aa3309e039d3940cc2; Jeremy Diviney/PR3148 f2927a4b8afcf0701a6b945e14f30de1d94f5176; PR3531 arithmetic 550a660a97308d3aeb3620ee1f4218f77e2c2742. Preserve the Yiyang Lu staged-kernel/flow, Nick Petalas copied-owner, Tekalign owner-checker, Marius-Paul Dumitrean endpoint/native, Jeremy Kothe preload/source, dataisfire/PR3361 bootstrap, and native PR2226 lineage with their original source notices and credits.

Prepared by Ross Richardson with substantial OpenAI Codex assistance. Separate AI review is not independent human review. The exact scope, proof, live inputs, freshly executed stages, explicit retained reuse and remaining hypotheses are recorded in the two packages.

[RossRichardson]

Submission hygiene review of 3efbb5de06a49ffeea6684482f2d3fe48cdca3fb: no submission-hygiene blocker found.

The PR title and description state the changed construction, conditional κ claim, exact sources, predecessor credit and substantial AI assistance. They distinguish the completed original release, the new wrapper's preparation-only execution, retained reuse and inherited mathematical assumptions. The canonical Discussion #3606 and its #1167 pointer were consistent with the reviewed head.

Source pins, artifact hashes and declared input paths checked cleanly; the diff is confined to the two declared packages. Automatic run 39113623065 completed both bounded package replays and their unchanged-input checks on this exact commit.

The committed validation JSON files are historical evidence snapshots. The PR and discussion provide the later publication and CI status. No README or validation-JSON changes are requested by this submission-hygiene review.

Reviewed with OpenAI Codex assistance. This is submission presentation/provenance review, not independent human mathematical acceptance or a fresh full release replay. The disclosed portable-replay follow-up and inherited all-size obligations retain their stated scope.

[thakrarsagar]

Note on the owner auditor at research/staged-644-660-conditional-release/owners/independent_literal.py (head 3efbb5de). It comes from an independent check of PR #3633: PR3694, finding 1. No correctness issue is involved.

The file's five reject() controls cannot fail:

• three predicates are unsatisfiable: a shifted column compared with the tile, lo<lo, and len({f,f})==2;
• two test a +1 perturbation of an equality that the preceding gate has just asserted.

So the controls never exercise the gates, and they pass unchanged if a gate is disabled. A local run of the T488 copy confirmed this.

The gates themselves are sound on the real data, and the package re-checks them with controls that feed violations. The docstring cites an original (SHA256 1001cb1c…) that I could not find in main or the open heads.

The same five controls appear in 15 files in total, including the copies in #3624, #3626, #3630, #3633, #3638, #3642 and #3653. Rerunning a factored check() on mutated journals, as research/rank7-joint-endframes/allocation/vendor/independent_owners.py on main does, would make them effective.

[RossRichardson]

Confirmed the owner-control finding at 3efbb5de06a49ffeea6684482f2d3fe48cdca3fb. The five controls do not invoke the all-row validator, and disabling the corresponding gates leaves those controls green. My earlier submission-hygiene comment predates this finding; it should not be read as resolving it.

The bounded repair is to factor the unchanged all-row checks into one callable validator, pass actual mutated journals/profiles through it, and require the intended rejection reason. The same inherited issue exists in #3630. Local patch candidates are prepared, but neither PR head has been changed, so the finding remains open.

The 1001cb1c… SHA256 identifies the preserved pre-adaptation author-run audit, not a published Git commit. I located and verified those original bytes locally. The checked-in implementation is complete and does not read that original file at runtime; its docstring should make this provenance distinction explicit.

The substantive all-row gates are present. This review found a control-sensitivity defect, not an arithmetic counterexample or grounds to claim a changed conditional bound. Existing exact-head CI and historical release evidence remain distinct from validation of any future repair.