Conditional κ = 0.001388049083313897: fresh partial-own asynchronous whole-owner certificate

[chafreaky]

Combines #507's partial-own factory and 84 kernel pivots with #473's nonuniform order of complete invocations and a new whole-owner allocation. The conditional result is κ = 0.001388049083313897, with complex supplier b = 0.001389978441615608493590. The #493 bit supplier remains nonbinding. Updated #473 at 9a2e1d4a7739aa4439982d6baea9ffca66bb8ae3 independently reports the same exponent with a different literal allocation; this PR matches that number and adds the full two-platform rebuild described below.

Simply shrinking the 42 changed helper blocks inside the previous asynchronous assignment loses the gain. The new allocation rebuilds all 40,320 roles. Explicit per-pattern/call capacities and literal bipartite matchings produce 800 distinct physical owners per invocation; the independent controller reconstructs every actual event window, chart, spectator and remainder. No residual is split across physical families.

The resulting profile is {1:179496, 2:16128, 3:40258, 4:6048, 5:25007, 6:4608, 7:528, 8:1224, 23:1, 48:734}. Live/physical stock stays 7,680/8,258, rank deficit stays 2,496, and all remainders are paid. The complete scalar/Horner invoice, asynchronous prefix, two rational moment engines, six bootstrap levels, all 47 strict outer inequalities and rejection controls are recomputed.

The verifier rebuilds both suppliers from scratch, including #493's bit operator, the new factory and retiming, exact scalar and frame checks, guarded instance and whole-program physical checks, then the new asynchronous composition. Actual bit profile, invoice and program hash bind the assembly. It runs no optimizer and accepts no cached supplier receipt.

One portability repair is explicit. Upstream #507's full Linux run failed its frozen source-word hash because two discovery-only floating display fields differed in their last bits from macOS. Every other generated field matched. The derivative pins those two historical display values to the published JSON values; the original source and exact comparison are retained. The source-word and candidate-word hash assertions are unchanged. An unchanged upstream Linux full pass is not claimed.

Validation:

• Fresh macOS arm64 and Linux aarch64 runs passed all five wrapper stages and all 21 dependency stages, including their nested full supplier replays. Complete stable result files are byte-identical: cb91ddf9cfac48e6fc8c6b734e377dbf80ba8cccb1226408b48e38734d9e4be7.
• Both reproduce κ = 0.001363614743745701: partial-own factory with 84 kernel pivots #507's published exact-math hash 90fd7d912c39187a6402e76d83e36f1cf2880a20e565a2c7c782bf19c3cc1f6c. The new controller checks 3,599 actual charts and 19 rejected mutations; prefix 205 bits with 512 charged; row coefficient 16,467 < 20,161.
• All 346 manifest files and 326 copied Git source files were checked. Receipts, toolchains, macOS stage logs and detailed audits are in research/partial-own-asynchronous-validation/.
• A separate two-engine certificate bounds further scheduling of this exact local word and fixed architecture by b < 0.001448157050078593686264. This is not a bound on different words or algorithms.
• Added a Python 3.14 workflow for the same full verifier. GitHub CI status is separate from these completed local runs.

Run from the repository root:

python3.14 -m pip install numpy sympy
python3.14 -B research/partial-own-asynchronous-whole-owners/verify.py \
  /absolute/new/output --boost-include /path/to/boost/headers

The inherited all-size compiler, mutable-basis/common-label transport, #484 Lemma A spectator realization, restored-row, precision/tape and analytic-transfer conditions remain. This is a conditional exponent certificate, not an unconditional multiplication theorem, Lean proof or measured speedup. The disputed #448/#494 cross-family split interface is not used.

Credits: eumemic (#507 factory/kernel84 and #493 bit supplier); Dugongue (#505 asynchronous scheduling and admission machinery); Rohan Arun (#473 nonuniform order and pattern/call-capacity repair); DreamingOfClouds (#449/#484 gauge/kernel interfaces); Ali Romman (#495 retiming); chafreaky (#458/#502/#506 adapters and prior compositions), and all authors retained in the pinned notices. New allocation, literal matching, composition and portability repair prepared by chafreaky with OpenAI Codex.

[Mister-Meeseeks]

Thank you @chafreaky. This contribution is now retained on main in the scoped 100-PR review and contributor credits.

Contribution: Partial-own factory with a fresh asynchronous allocation.

Review and limits: All 40320 roles are allocated anew; simply shrinking an old allocation is not the certificate. Matching a parallel result still merits separate construction/replay credit.

The complete original source, notices and relevant execution evidence are retained through the pinned source archive. The review distinguishes fresh local runs, source-bound historical Linux CI and prior exact-head review; it does not claim 100 fresh full mathematical audits. Closing this historical PR as reconciled. This preserves parallel and cumulative credit without replacing the selected frontier or promoting priced research targets to completed constructions.

Reviewed head: 73d367fe754cf5921cd954221133b20ac752f932. Publication: 64039f1bd3b6a523b96083fd858ad040a75e3030.