Frame-filtered collective kernels + target-prefix sharing: kappa = 0.000712074434532274423885827

[Dugongue]

Conditional kappa = 712074434532274423885827/1000000000000000000000000000 = 0.000712074434532274423885827. Approximately 0.0943% above PR266.

Adds research/filtered-kernel-target120/, an offline source-bound finite construction and verifier. The public all-size interfaces remain assumptions; this is not an unconditional multiplication theorem or a Lean certificate.

Contribution

Filter helper columns by an actual common nondegenerate address frame before choosing a response basis. Expose the resulting simultaneous kernel directions, then select compatible blocks using chronological donor sharing. The resulting word has 925 pivots and entrance rank 1762, versus PR266's 851/rank1677. One selected collective component has eight pivots across 18 helpers at address rank 11. The bounded search is not claimed optimal.

Compose that word with PR268's 217 target-prefix groups, rechecking all 246 relations and actual frame paths. This removes 218 local recursive children while paying 275 scalar ADDs. Apply PR263's 25 retimings and PR261's fixed-prime arithmetic. Gains are recomputed from the complete word, never added as estimates.

Validation

Default python verify.py --output ../filtered-kernel-target120-replay regenerates pinned PR249 sources, compiles nine C++ checkers and executes the rational retiming stage. It checks all local formal columns forward/inverse; all 23627 five-stage columns; prefix relations, source ownership and COPY lifetimes; 1142 exact charts; 9952200 bank assignments; complete finite costs and payload bounds; two rational moment engines; eight ordinary levels; all 47 strict constraints and adjacent-grid rejection. Every expected receipt and transformation hash is pinned. Row swaps cost four primitives, and all selector costs are included.

Retained attribution, licenses, assumptions and assistance disclosures are in NOTICE.md and the proof notes. No personal byline is added. Higher unbuilt public targets are not treated as witnessed baselines.

[Mister-Meeseeks]

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

Contribution: Filtered collective kernels and later gen4 endpoint-aware composition.

Review and limits: Fresh full gen4 endpoint-composition replay passed. Complete package comparisons bind the older chain to prior reviews and PR267/242/271 checks; the 48-file filtered parent is exactly the one used by PR271. Historical auxiliary snapshots retain their separate scope.

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: 65768f62d4fde0d1fa7fc84f71678522a84aaa6c. Publication: 50d768818f579d4e165b4c257c6ee25bca300802.