Skip to content

Use the minimum physical odd scale in the interleaved construction - #7730

Open
aliromman wants to merge 1 commit into
CrocSwap:mainfrom
aliromman:codex/consecutive-result-058
Open

aliromman wants to merge 1 commit into
CrocSwap:mainfrom
aliromman:codex/consecutive-result-058

Conversation

@aliromman

@aliromman aliromman commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor

The exact interleaved graph can use native odd scale 176,715 in place of 2,650,725. With binary height 21 and the source root format retained, the complete ordinary bill lowers its conservative scalar-service rate from 2^30 to 2^26. This is a factor of 16 reduction in one bound coefficient. It is not a measured runtime or total-cost gain. No new kappa follows; every paid and productive job is unchanged.

This enabling proof resolves the physical-scale question in PR #7726. Read the proof, complete receipt, verifier, source pins and credits. The source already reported the normalized minimum product grid 723824640. This PR binds its odd part to physical components, per-call cleanup, the common root and the complete finite bill.

Check Result
Every separate coefficient product and public endpoint 10,483,566 products; all 1,260 endpoints
Sharp physical necessity Explicit witnesses for 3, 5, 7, 11 and 17; every scale missing one required prime fails
Scalar implementation 1,108 factors; 6,648 signed cases; maximum coefficient bits 22 → 18
Complete 16-call invoice 96,512 paid jobs, 17,296 productive jobs; unchanged
Ordinary charges 123,035,648 scalar services, 47,392 odd conversions; all counts and layout formulas unchanged
Physical storage All retirement and guards restored; live-sender overwrite and partial-clear controls reject

The exact graph is ba59d5db17c40185c8608a734ced10135d8c041e432c3840e9034868781c31a2. The universal minimum uses arbitrary independent native inputs and the inherited complete Gaussian-dyadic chart contract. It does not assert a minimum for a restricted correlated caller. The storage fixture has four addressed cells per home; whole-extent behavior remains inherited. The source constructor and physical/caller engines are shared dependencies.

The general scale theorem and verifier structure come from PR #7718, pinned proof. Retain Gareth Davidson / bitplane's grid criterion in PR #866, Ali Romman's PR #7680 coefficient witnesses, and the common-root proof. The retained notices preserve icekylinx, Clear1nsight, DreamingOfClouds, SYLforge, Rohan Arun, Jerry Chen / CubeJerry, Douglas Colkitt and all predecessor credit.

The source scientific commit is 3ba9094c54b61f424d2f98cd2c6ca64c01baf9d3, published at 925719c8ae6d43216ff1f2a87c8da15f12525081; its source credits identify the construction and Codex assistance. This PR starts from canonical main e54523517d0f03e58352e647c403a5af3e525a57. All 63 retained files are byte-pinned. The pending source commits are copied dependencies, not asserted publication ancestry.

Python 3.12, standard library only:

python3.12 -B scripts/research_package.py --package research/interleaved-minimal-odd-scale

Both normal and optimized runs print:

PASS_MINIMAL_ODD_PHYSICAL_SUBSTITUTION 176715 16

Normal and optimized replay, commit-byte/source-history/mode checks, export verification and a read-only sealed serial replay pass. The exported replay used no Git and left all hashes and modes unchanged. Separate rational columns and actual physical chart amplitudes check the five prime witnesses. Contract controls cover nonunit charts, correlated input cancellation and a unitary phase with an odd denominator.

The five native/owned/compiler/precision/analytic interfaces remain conditional. The original root bound and compiler H remain conservative. Full phase catalogue, compiler search, historical suites, selected cold replay, Lean and Docker were not run. This is author verification with reused engines, not blind, independent human or formal-kernel acceptance. The selected frontier is unchanged. Hosted CI did not certify this package. The plan job failed in the shared test_current_ci_section_matches_selected_record: the base guide names frontier6030 while the selected record requires make frontier7000-verify. The log reports 760 tests, one failure and 24 skipped; package replay was skipped. This package changes no shared CI file. Local replay remains the scientific evidence.

Prepared by Ali Romman with substantial OpenAI Codex assistance in research, implementation, verification and writing. Apache-2.0. Canonical claim and handoff.

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.

1 participant