Repository navigation
test(research): certify the V45 closed native endpoint cone - #3038
Merged
Mister-Meeseeks merged 1 commit intoOct 16, 2026
Merged
Conversation
Source-complete two-provider/two-orientation causal cone audit; exact native class checks, independent SSA closure, behavioral controls and cold replay. Substantial OpenAI Codex assistance. No new supplier or kappa claim.
This was referenced Oct 17, 2026
Mister-Meeseeks
pushed a commit
that referenced
this pull request
Oct 17, 2026
…the T74 invoice (no new kappa) Checks PR #2979's mixed-entrance group A (helpers 331, 450, 451, 508 and 509) on both literal providers in both orientations. Helper 331 enters cold and its emitted residual 0:261 has width seven; the four helpers entering at [30] use 115:261, of width six rather than the terminal dimension seven. Both residuals hold phase-exactly on all 512 address columns. Each helper's complete direct path (scalar clocks, climbs, private lifetime and both wrappers) is bound to the literal source, all ten helper/private endpoints are normalized identities under the retained PR #3006 engine, and the widths are joined to all 3,330 group-A placements of the final T74 allocation. A time-major recount of all 271,728 roles and 149,184 bins derives the idle bill 17310 H1 + 4956 H3 + 1728 H5 + 28190 H48 and the complete bill {1:1645680, 2:470196, 3:159468, 4:71928, 5:13716, 8:5328, 48:28190} (rank mass 4,816,512 = 65,088 x 74); nominal stock is 66,813. PR #3038 already checks both residual classes in its V45 cone, so the native identities are corroboration; none of the five post wrappers is in that cone. The PR #2993 group/control replay (joint delta -H1+20H2-14H3+2H4-2H5, guide margins +42,541 and -58,943) is credited reuse. No defect was found. Peer-native and warp closure, the other 86 private endpoints, physical stock and movement, prefix/precision/tape, legal control allocations, both supplier caps and the outer and all-size premises remain open; no supplier admission or kappa change. Adds research/rank7-group-a-entrance-audit/ only. Submitted by @commdata2338 with substantial AI assistance (OpenAI Codex research and implementation; packaged with Claude Code). Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01X86sCSXdoDd61SDFfC1PKG
Mister-Meeseeks
pushed a commit
that referenced
this pull request
Oct 17, 2026
…dependencies Answers PR #3038's first open question for the original PR #2979 words. In both providers and both orientations, all 91 private buffers are zero immediately before their actual frees for arbitrary original source, target and dirty-helper inputs (364 checked frees), by exact rational propagation over reduced noncommuting warp words. Forward value, chart and lifetime provenance seeded at every free enlarges PR #3038's two-root cones per provider from 8,902 to 12,693 events forward and from 12,357 to 20,956 in inverse transpose; all 5,815 expanded native classes per provider (609 and 1,377 newly reached) pass exact 512-column identities including global phase. Flipping cleanup event 17256 (Z395 -= L5) leaves PR #3038's helper249/Z249 roots correct but leaves Z395 = -2(x64+x65+x96+x97) at its free, event 22501; the dual flip, premature frees and a newly reached native phase mutation also reject. All 91 private families stay charged (peak 91, stock 66,813); the complete invoice is reconciled arithmetically, not re-derived. Conditional on the normalized macro/native boundary; physical scalar/COPY/CLEAR, backing ownership, precision/tape, both supplier caps, the outer release and all-size conditions remain open, and no supplier, kappa or selected-frontier change is claimed. Adds research/rank7-peer-private-free-audit/ only. Submitted by @commdata2338 with substantial AI assistance (OpenAI Codex research and implementation; packaged with Claude Code). Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01X86sCSXdoDd61SDFfC1PKG
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
2 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
This finite conditional audit reconstructs the complete frozen PR2979 emitted source sequence, then checks the closed value/chart/lifetime dependency cone for helper249/private953 and every native class it reaches, for both final providers and orientations. The normalized helper endpoint is its own original input and private953 is zero immediately before free; the paired physical wrapper factor is F7/inverse F7.
Per provider, the cone reaches 8,902 forward and 12,357 inverse-transpose events out of 22,748. These include 2,819/2,601 native events in 2,039/1,790 classes. All 7,658 class checks include all 512 columns and global phase, with every event associated to an exact identity receipt. Two independently implemented closure algorithms, reverse demand and forward SSA, agree against handwritten normative controls.
Physical rational scalar/COPY/CLEAR and known-zero allocation contracts, every peer-private return/reuse obligation, physical ownership/final packing/storage/precision/tape, and whole-supplier/outer/all-size/compiler admission remain open. No new supplier, cost, κ, selected record, independent human review or proof-kernel theorem is claimed. Only one isolated research package is added.
Task and overlap
Task
independent-moment, priority revision 6, persisted seed 78; scope refined. Based8e0f83f00191df62174b641e9fff055c357f641; frozen construction PR2979 atbe191346391661f4d35fc0d1508d0a6029572576.The initial joint-group/single-change and fixed-parent invoice slices already overlap PR2993/3014, so this takes PR3008's released full-peer dependency handoff instead. PR3006,
1432a43a331ec857f45093a1d6e16acdbe9849e2, checked selected normalized endpoint covectors and seven native recipes. PR3008,948a997eba1aaf13d04e2b52edf70404860d19dc, bound 73 directly touching events. This package independently reconstructs the full source sequence and adds the complete causal peer/native chart closure, with full-payload orientation-aware caching. It imports neither predecessor checker; the frozen producers and credited separate exact kernel are reused.Canonical Discussion2612 and the latest100 updated open PR bodies were refreshed, followed by the latest30 all-state PRs through3036 immediately before publication. Main remains at the base pin. Later word/packing compositions are separate and are not independently admitted here. This bounded coverage does not guarantee novelty against unpublished work or arbitrary diffs.
Validation
Exact head
c7d0d7be056ba77fda1c168641b2bcffb10cf43d, tree952218dc7ccd0334a4c7c82c4f14f6452c3ce939; 17 files, all underresearch/rank7-v45-native-cone-seed78/.Supplied-artifact all-class checks passed, followed by the unmodified registered default cold replay in 599.18 seconds total. Both providers reproduced the committed deterministic certificates. CPython3.12.14/NumPy2.3.5, isolated optimized warnings-as-errors checker interpreters, no dependency installation. Metadata/Git pins, source/code stability, AST compilation/no assertions/function size, staged whitespace and separate internal static review passed. VALIDATION.json distinguishes actual execution modes and retains receipt/code/source hashes. Only explanatory documentation and the validation record changed after cold replay; checker/source/certificate bytes are unchanged.
Complete source binding checks exact warp matrices, center containment, FULL cleanup and private lifetimes. Semantic mutations bypass hashes: an actual peer warp crosses an ADD while the old direct projection/clocks stay unchanged; wrong peer native phase/character rejects with an explicit Gaussian basis witness; conjugation differs from adjoint. The same-rank center control is explicitly a derived center-only fixture. Eight normative graph positives, a missing-edge mutant and fifteen oracle input guards distinguish closure agreement from decoder correctness.
Full selected-result cold replay, Lean, Docker and hosted CI are unclaimed; Ruff was unavailable locally. Finite Python evidence and a conditional composition argument are not a full multiplication theorem.
Credit
Credit Marius-Paul Dumitrean for PR2979; Nick Petalas for PR2993/3006/3008 and the released handoff; the retained transport/native/extraction authors and Sagar Thakrar's separate exact kernel. Full lineage and source/license pins are in README/SOURCES, including the retained PR2226 notice. ToikY with substantial OpenAI Codex assistance, including a separate internal oracle implementation and static reviews; no independent human review. Apache-2.0.
Open next questions
Package follow-ups: seed all actual peer-private free obligations to check ancillary reuse, then join the local execution with PR2987's actual final fill and complete physical invoice. A peer's later free need not be a dependency of these two chosen roots; scalar realization, backing ownership, precision/tape and whole-supplier contracts still block admission.