Repository navigation
Conversation
|
Thank you @eumemic. This contribution is now retained on main in the scoped 100-PR review and contributor credits. Contribution: Kernel entrances with 217 target-prefix groups. Review and limits: Source-bound full Linux replay of the combined word. Keep dirty responses, chronology and bank provenance bound to the same transcript. 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: |
2 participants
κ = 7.11191385398487e-4
Conditional κ ≈ 7.11191385398487 × 10⁻⁴, from a fully replayed composition of PR251's 37 transported entrances, PR254's 132 response-kernel pairs, and 217 new target-prefix groups covering 872 targets. This is a fresh branch and self-contained package on main; prior PRs remain frozen.
For each group, the verifier derives the exact F₂ dependence of target-prefix responses from the actual word. Zero-frame setup removes the dependent response, and restoration at a checked common frame preserves arbitrary initial target contents. This removes 1,081 prefix reads and inserts 1,316 setup/restoration additions. The complete resulting word has 786,594 weighted ADDs; every source, target and dirty column is checked in both directions. Both reflected ledgers are rebuilt from actual uses. No savings are added arithmetically from separate candidates.
The literal 120-replica construction has stock 2,606,465, 58,900,800 paid children, rank mass 312,247,800 and deficit 528,000. All 9,952,200 bank assignments and 377 exact charts are regenerated. The full finite invoice charges 1,949,778,076,800 additional bank selector calls, including the new charts' 576-factor maximum. All other routing, copies, boundaries, cleanup and fallback remain charged.
Validation: two independent complete
python -Breplays returned exit 0, all eight mandatory stages passed, the 135-file manifest was unchanged, and the certificates were byte-identical. Checks include all 20,107 local F₂ columns and literal inverse, signed source527 decoder/restoration, actual five-stage lowering, both reflected frames, all used-basis prime exclusions, bank endpoint controls, two exact moment engines, 47 strict assembly constraints and adjacent-grid rejection. The certificate and both replay reports are included undercertificates/.Requires Python 3.11+, SymPy 1.14.0, assertions enabled, and a new output directory outside the package. Replay needs no network. A dedicated workflow runs Python 3.11 and 3.13; no CI success is claimed before it runs.
The exact claim is 711191385398487/10^18. At the final pre-publication check this exceeds PR265's κ ≈ 7.11032242134072 × 10⁻⁴. PR262 explicitly describes its higher value as a priced target with open construction obligations.
The retained all-size compiler, common weighted charts, restored rows, selectors, routing, prime supply, precision/recovery, complex symbolic correctness and analytic transfer remain conditional interfaces. F₂ payload algebra is kept separate from odd-prime address arithmetic. This does not claim an unconditional multiplication theorem, a practical runtime improvement or a new Lean proof.
Provenance: Dugongue's PR254 supplies the 132-pair kernel construction; eumemic's PR251/249/244 supplies the transported entrances and source retiming. The package retains PR210/200/193/207/211/216/230 lineage, hcg890/Jacob Sussman's PR234 five-stage architecture, Evan McKinney's PR197 bank method and Rohan Arun's PR237 width-120 application. Original notices and licenses remain included. Target-prefix compression, composition, checks and packaging were prepared by eumemic with substantial OpenAI Codex assistance.