Skip to content

Prove exact phase entries for 2,304 denominator-52 tails - #7757

Open
aliromman wants to merge 1 commit into
CrocSwap:mainfrom
aliromman:codex/consecutive-result-071
Open

aliromman wants to merge 1 commit into
CrocSwap:mainfrom
aliromman:codex/consecutive-result-071

Conversation

@aliromman

@aliromman aliromman commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor

Exactly 2,304 of the 2,704 actual denominator-52 residue pairs have a unique entry into the existing nonzero denominator-13 arithmetic core after at most four coins. This supplies the missing parameter, phase and live-node bindings for that part of the PR #6096 direct-graph question. It also proves that the first-entry inverse needs a separate boundary protocol. No new kappa or physical entry tariff is claimed.

Read the proof and exact contract, entry formula, complete pair certificate, checker, 29 source pins and credits.

The contract keeps the original order 00,01,10,11, every unfinished node and the ordered coin suffix. For 13 dividing neither a nor b, the unique core parameters are c = 10a mod13, d = 10b mod13, with phase theta = ab/(cd) mod169. The 2,304 inputs use 1,560 distinct recipes. A separate inventory reconstructs all 22,464 core signatures from literal source transition labels and agrees with the formula.

Actual input class Pairs Tail after four coins
Neither residue divisible by 13 2,304 Unique entry into the current core
One residue zero, the other not divisible by 13 96 Endpoint tail outside that contract
One residue in {13,26,39}, the other not divisible by 13 288 Mixed tail outside that contract
Both residues divisible by 13 16 No unfinished tail

The exclusion concerns this complete-pool interface only. It does not exclude other controllers, endpoint dispatch or state encodings. Early leaves use fewer than four coins; (0,0) uses none. An unfinished pool must not be replaced by a fresh one-root sample without a proof.

All 2,704 bounded owner witnesses, 43,264 literal four-coin prefixes and 1,437,696 eligible tail transitions pass. The entry-specific inverse passes for all 4,428 continuing entry nodes. At the actual source pair (1,1), prefix 1111 enters core phase 120, node zero; its prefix predecessor is phase eight, node zero, while the core cyclic predecessor is phase 60, node one. A transition that merges both inputs into the same output and clears its workspace cannot be reversible. A separate entry/return protocol can work. No universal extra-bit lower bound is proved.

Python 3.12, standard library only:

python3.12 -B scripts/research_package.py --package research/inverse52-phase-entry

Both registered commands print:

PASS_INVERSE52_PHASE_ENTRY
2704 actual pairs; 2304 pairs with a unique core entry; 96 endpoint tails; 288 mixed tails; 16 dyadic pairs
22464 backend phase signatures; 43264 literal prefixes; 1437696 tail edges
4428 entry inverse checks; explicit first-entry/period-return collision

Normal and optimized registered replay, syntax/indentation, committed bytes/source history/modes, export and sealed read-only Git-free serial replay pass. The complete unchanged PR7755 verifier also runs, including its circuit and fixed-pair checks. 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.

@JoyboyBrian supplied the actual residue image and nonadaptive arithmetic window, domain checker and direct-graph question in PR6096. The exact mode columns are retained. Ali Romman (@aliromman), with substantial OpenAI Codex assistance, supplied the runtime core and replay in PR #7755. Its whole package is retained byte for byte.

That package credits Nick Petalas (@petalas) for the PR #7746 arithmetic primitives, @zxv for the PR #6072 paired graph and probability proof, and @JoyboyBrian for the PR #6102 inverse/lifecycle context. All retained notices and predecessor credits remain. The numerical caller in PR #6546 is a distinct result; this PR does not claim that numerical caller joining is globally absent.

Physical prefix execution, parameter loading, entry/return control, wire/tape access, stopping storage, output retention, conditional freshness and the complete native invoice remain open. The complete source-domain kernel fixtures, historical probability/mean catalogue, native/selected cold replay, Lean and Docker were not run. All five framework assumptions and the selected frontier remain unchanged.

Base: canonical main e54523517d0f03e58352e647c403a5af3e525a57, priority revision 8. Consumed heads: PR6096 e56f5094c83f3e514d9936fa7b1c8027745a0a00 and PR7755 d6babec41a3d01928c3dbdb896ea2adf051ba305. No unmerged source ancestry is asserted. Prepared by Ali Romman with substantial OpenAI Codex assistance in research, implementation, verification and writing. No independent human or formal-kernel acceptance is claimed. Apache-2.0. Canonical claim.

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