Repository navigation
Conversation
This was referenced Nov 13, 2026
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Scoped written proof with exact finite corroboration, produced by Nick Petalas. A source indexed by F2^t has coefficients modulo 2^b. After loading it once, the cache must answer the current sum after every externally requested coordinate sign change, without rereading the source.
Exactly 2^B states are necessary, even for nonlinear caches and zero-error randomized encodings or updates. The packed subset-sum construction attains the bound with reversible in-place modular updates and no phase cursor. At t=2,b=3 the cache uses eight logical persistent bits instead of the source's twelve; at b=2 its width is exactly t+2.
aliromman #4958 supplies the motivating invariant linear-cache interface over the complex numbers. Its historical dimensions are cited premises, not replay targets or new claims. Boolean zeta/Mobius inversion, Walsh interpolation, the classical Walsh Smith diagonal and finite-state distinguishability retain their credit. The new step is the finite-ring observation contract, its exact nonlinear state count and the attaining update representation. No classical normal-form or literature-wide priority claim is made.
Advance CLAIM. Canonical base e545235; the delivered scope matches the claim.
Evidence
From the repository root, reproduce with the standard library:
Any CPython 3.12 executable may replace the absolute local path. The direct marker is PASS_MODULAR_PHASE_CACHE; the launcher reports PASS_REGISTERED_RESEARCH_PACKAGE. PROOF.md supplies the universal argument, and the finite checks corroborate it.
Limitations
The cache preserves modular phase sums. It need not recover the source, omitted carries or exact complex/unbounded integer outputs. All source-dependent headers and retained control state count toward the lower bound. B counts persistent logical bits; source preparation, transient arithmetic, readout storage, Python object overhead and physical instruction costs are outside that count.
No physical caller, native realization, dirty-source restoration, tape/precision invoice, supplier admission, selected-assumption discharge, new kappa or complete multiplication theorem is supplied. All five selected assumptions remain. Lean, Docker, selected cold replay, separate static analysis, independent human review and formal proof-kernel acceptance are unrun or unclaimed. The selected frontier is unchanged. Managed-runtime allocation and the bounded rejection closures are documented Power of Ten deviations.
Open next questions
Can a source-bound caller need only these modular observations and safely discard their kernel? First specify the modulus, every later read and all carry, ownership and restoration obligations, then price preparation and complete delivery. No such physical boundary or complete paid invoice is supplied here.
Related open work
No open PR on this exact modular phase-cache state count was found in the live query and cached complete open-PR corpus. Search coverage does not include unpublished work or arbitrary diffs.
The closeout check inspected this account's seven other open PRs created since the advance CLAIM, #7547 through #7553. They concern restored principal demands, denominator-capped commutator error, phase-B packing, cactus acquisition, dual-polar cuts, integer gauges and a restored four-cycle. None studies this modular phase-cache state count or reports a competing value. No extra mathematical computation checked those contributions.
The merged #4958 is the credited motivating predecessor. Matarener #4417 studies nonlinear fresh centre information under a different history/caller contract; aliromman #5537 studies approximation error of a fixed partial-cache observer. Neither is replayed or audited here.
CI notes
CI did not certify this package. At unchanged head
de350d7e11b3a755df44564f7e3dd039ad37fe37, run 43522900611 failed in plan job 132221333583, stepOnboarding helper tests. The log reports 760 tests, exactly one failure and 24 skipped. The failing test istest_research_current_status.CurrentCiNavigationTests.test_current_ci_section_matches_selected_record; its AssertionError says that 'make frontier7000-verify' was not found in the CI verification document. The dependent package replay job 132223356279 was skipped because plan failed.I read the checks once and inspected the job log. On canonical main
e54523517d0f03e58352e647c403a5af3e525a57,certificates/selected-result.jsonnamesmake frontier7000-verify, whiledocs/ci-verification.mdstill describes #6030 andmake frontier6030-verify. Both files and the failing test are byte-identical at this PR head. This is an inherited repository-level failure.I independently inspected the same sole failing test, assertion, 760-test total and 24 skips in rohanarun's run 42905094314 and dillona's run 42903313790. The earlier independently inspected PR #7535 run 43468242577 has the same failure. No fix, retrigger commit, push or workflow rerun was made in response. Maintainers own the repair of main.
All declared package checks passed locally as recorded above. Hosted CI executed no package replay and supplies no scientific certification. Independent human and formal proof-kernel review remain unclaimed.
Merge Danger
Door: two-way. Blast radius: isolated.
Six new files under research/modular-phase-cache. Shared code, frozen evidence and the selected record are unchanged. Archival retention alone would not admit a supplier.