Repository navigation
Owner consolidation on PR610: conditional κ = 0.001625434703655052 - #618
Merged
Mister-Meeseeks merged 9 commits intoOct 12, 2026
Merged
Conversation
…er call via commuting-prefix cleanup
…on the PR601 witness (+0.368%) research/complex-571-cold14 starts from PR601 (jeremykothe2, complex-571-hpem46) and appends one typed build step, cold_end, to each provider chain (ck, atm3, jp8): in every call 14 cold factory-scratch helpers (S b.10) end at a 4-dimensional frame instead of FULL (width 9 -> 4) under a refactored exact cleanup (rank-0 dirt-spread premix at the head of A, 56 multi-gate lower-frame tail gates, the 46 inherited hyperplane-end gates, and an exact dyadic FULL remainder). Dispatch ck 44 / atm3 22 / jp8 6 and the literal owner map are re-solved for the new widths. Verifier: cold helpers may end below FULL with (0, U) wrappers (exact Lemma A check per pair), a refactored-cleanup switch (exact replay both orientations, dyadic, non-target), a row/column cone pre-check, and wrapper-order checks at both the transport and the event level; 14 raw-word mutants for the new step. Builds on PR591/PR592/PR593/ PR596/PR600/PR601 (jeremykothe2), PR531 (icekylinx), our PR509/PR516/PR599. Build on this at your own risk. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ude Opus 5.5 via Claude Code) Attribution line made exact; PR601's own attribution sentence is quoted rather than replaced, and the inherited owner-map sentence now says it describes PR601's map. MANIFEST updated. No code, data or kappa change. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…1-cold14 (+0.439% over PR601)
Same providers, cold_end steps, verifier and dispatch (ck 44 / atm3 22 / jp8 6); only the literal owner map
(data/OWNERS.json.gz, data/PLAN.json) and the recorded expected values in data/SPEC.json change. The map
comes from an exact pattern-restricted owner model (one helper per port per call, CP-SAT warm-started from
the annealed map) expanded literally by bipartite edge colouring; the package verifier re-admits it
independently. Paid profile {1:185792,2:32576,3:23600,4:14128,5:5040,6:2016,8:576,32:1,48:2094}.
README: new kappa, owner-map method, remaining headroom ~0.06 pt on this dispatch, and the jp8 note
(time-refinability does not bind at the LP level; the jp8-heavy gap is integrality and search).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ct integer owner map for complex-571-cold14 (+0.252% over PR604) Stacked on PR604 (1b795c5, kappa 0.00162128508944984). Providers, cold_end steps and verifier code are unchanged; the dispatch moves from ck 44 / atm3 22 / jp8 6 to jp8 71 / ck 1, and the literal owner map is re-solved for it. The map comes from a per-owner-group integer model (CP-SAT) over width patterns with the one-helper-per-port-per-call constraint: on the all-jp8 dispatch it is solved to optimality and equals the LP bound (the time windows and time-refinability do not bind), and one ck call is added by a warm-started repair model. The integer solution is expanded literally by bipartite edge colouring (Konig) and re-admitted independently by the package verifier. The audit needs two dispatched providers (substitution controls), so a pure jp8 dispatch is not packaged; the atm3 chain is no longer dispatched and its recipe files are removed. data/SPEC.json expected values are re-recorded (31 -> 23 audit mutants, 155 -> 114 assembly controls with two providers). kappa +0.252% over PR604 and +0.692% over PR601. README describes the change, the remaining gap to the free-dispatch LP (about 0.026 pt of b), the certified scope of the 14 cold units, and other provider words. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…bjective note (review fixes) Text only: the native complex lower bound now matches ASSEMBLY.json exactly, the README says the ck and jp8 words are unchanged and atm3 is removed, the owner-map table notes that its values are rounded CP-SAT objectives, and the inherited NOTICE sentence is marked as describing PR601. MANIFEST updated; no data, code or kappa change. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Change and mathematical scope
Consolidate twelve helper placements in PR610's literal owner map, raising its conditional κ from 0.00162536523418317 to 0.001625434703655052 (+0.00427408%). The conservative value is 0.001625434700837327. The dispatch remains jp8 ×71 / ck ×1; every provider circuit, construction recipe, external source pin and verifier script is unchanged from the pinned PR610 head.
The repair in PR610 leaves several rank-48 owners with one helper. This change moves helpers from different complete calls into fewer owners while preserving the idle windows, capacity and one-helper-per-family-per-call constraints. Twelve records change placement and two additional records change only packed offsets. The standard-library discovery tool reproduces the frozen map and records all twelve moves; its floating-point search objective is not trusted for admission. Exact κ is
406358675913763/250000000000000000.Dependency and review scope: this is stacked on #610 at
b7c680ea38b9a25a89fce189bb927b45842362d0, which builds on #604 and #601. Those inherited commits appear in the comparison against main while the dependencies remain unmerged. The numerical change and nomination tool are isolated in commit c0874b2 (14 files); subsequent updates add validation evidence only. This does not update the repository's selected certificate.Concurrent results: #614–617 and #622 now report stronger κ values. This submission contributes a small reproducible owner-consolidation tool, a verified improvement over pinned PR610, and its replay evidence; it makes no current-record or optimality claim.
Detailed changes and reproduction commands: OWNER-CONSOLIDATION.md.
Validation
make gen4-bank-verify: all nine stages and 112 immutable source pins pass. Full repositorymake verifysubsequently passed on submitted headc0874b2, including all 84 isolated test modules and historical patch checks; the checkout remained clean.Retained replay evidence contains the result, stage receipt, assembly controls and contributor validation record. This is contributor validation awaiting independent review. The cited PR543 bit supplier is repriced but not rebuilt; the inherited all-size, compiler, hosting, precision and analytic assumptions remain. No unconditional multiplication theorem or full Lean certification is claimed.
Credit remains with DreamingOfClouds (Claude Opus 5.5 assisted) for PR604/PR610's cold-ended providers, dispatch and owner construction; jeremykothe2 and all predecessor contributors retain the detailed attribution in NOTICE.md. This owner consolidation was prepared by Jeremy Diviney with OpenAI Codex assistance.
Follow-up #624: our newer owner-repacking contribution builds on the 492-helper construction in #622, with conditional κ = 0.001635425470780892. It also documents the stronger concurrent #623 result.