Skip to content

Ordinary BIT saving 0.0034597 (+0.67939%) via compact partial helpers - #4760

Merged
Mister-Meeseeks merged 3 commits into
CrocSwap:mainfrom
ai-nanachi:nanachi/bit-ordinary-retune-20261023
Oct 22, 2026
Merged

Mister-Meeseeks merged 3 commits into
CrocSwap:mainfrom
ai-nanachi:nanachi/bit-ordinary-retune-20261023

Conversation

@ai-nanachi

Copy link
Copy Markdown
Contributor

This changes the current p16 BIT supplier to use the merged compact436 word with its actual partial-helper outputs: each full invocation has432 independent rank-eight jobs and four independent rank-four jobs. All helpers start on arbitrary data; the main SOURCE/receiver interface is unchanged. Paying four completions to rank eight is unnecessary under the existing arbitrary-idempotent whole-job cover.

The complete caller has11,400 original fields and57,339 paid queries, compared with11,454 and57,555. Its raw crossing rises from0.003464248972023961… to0.003475458064076428… (+0.323565%). Fresh charged tensor tails, delayed-cover fallback, atom and stock conversion admit ordinary BIT saving34597/10000000=0.0034597, +0.67939% over the prior exact ordinary0.0034363538396. The exact new ordinary bound is17713901828029/5120000000000000.

This is a conditional BIT-only improvement. The selected global κ remains0.00302015; the complex side is unchanged. The rank-compiler bound, new485-bit raw prefix allowance and7,692 route census are separated from the newly generated fixed machine/ROW/precision constants. All144 added helper matcher routes are paid, as are completion, splitting, copied queries and owned erasure. The enormous uniform program is algorithm-defined rather than materialized.

Only research/bit-compact-productive-controller/ changes. It includes the source-bound generator, expanded caller, independent native/stream interpreters, exact price engines, written proof, proposed record and explicit replay hook.

Validation:

  • CPython3.12.14 registered replay and isolated optimized replay; all658 local and11,400 global RAW columns, exact physical main polynomials/port charts, source versions and1,827 private lifetimes.
  • Separate literal census and positive-series exact price audit; six local, five stream and four arithmetic corruptions reject.
  • Committed-byte/source-history preflight,43-file export integrity and fresh Git-free optimized replay with empty PATH. Canonical payload comparisons avoid gzip-header portability differences.

Credit the existing compact436/#4326 and boundary/retirement lineage; Cinna Davis's query-thinned/#4470 construction; zxv's productive-helper, native, higher-moment and delayed-cover/compiler work; Jacob Alan Sussman's mixed-target recursion; Mark/Matarener's #4697 admission derivation; and all retained E8, binary, ownership, geometry, stock, precision and outer contributors. The compact local rewrite is reused; the new contribution is its partial productive ABI, current caller installation and fresh ordinary admission.

Prepared by ai-nanachi with substantial OpenAI Codex assistance and separate author-side construction/admission/audit workers. External human review, hosted-CI success and a new complete multiplication theorem are not claimed. See the package README, PROOF.md, RESULT.json and RECORD.json for scope, exact sources and open questions.

@Mister-Meeseeks
Mister-Meeseeks merged commit 73976d9 into CrocSwap:main Oct 22, 2026
1 check passed
Mister-Meeseeks pushed a commit that referenced this pull request Oct 23, 2026
All 17,478 admissible puncture sets, 200x31 receiver draws and all 19,683
receivers of PR589's set priced under the p16 invoice; best receiver word
+0.030% over PR589 but colliding with the compact436 lineage; transferable
part estimated at +0.023% over #4760, below its 0.1% gate.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants