Repository navigation
Certify the staged544 single-ADD retiming obstruction and a paid circuit control - #9
Conversation
… control Add an isolated exact census of both PR3734 staged544 providers, an all-exponent non-improvement proof for the declared retiming class, and a credited PR2957 rollback/recovery control with PR3863 paid credit. No new supplier or selected-frontier change. Prepared by Chris R / Clear1nsight with substantial OpenAI assistance. Retain all source and interpreter attribution.
Carry the focused change from 89031d2 (PR #8) onto this active branch. Initialise TMPDIR after runner setup, preserving workflow triggers, repository guard, permissions, pinned actions, timeouts, replay commands and all research inputs. Validated YAML, exact semantic delta, shell syntax and TMPDIR assignment including spaces. Substantial OpenAI assistance.
|
FINDING / DONE / 2026-10-19 04:34 UTC. The user-requested neutral-first follow-up is published as fork PR #12, #12, at 1416f8c. Only research/staged544-local-two-add/ is added; frozen predecessor evidence and both default branches remain unchanged. The completed source-bound result covers any radius-one or stationary first ADD with every legal first frame, including locally losing first moves, followed by an unrestricted second ADD inside its allowed corridor. Both providers have 5,756 nontrivial first-move records and 11,917,718 reduced second price cases, with 49 exact non-improving signed types for every real 0<a<1. New interaction/new-gap reductions and independent finite controls make the scope reviewable. Both full local replays and normal/-O controls pass. The complete published 20-file subtree matches the tested local tree, 2475d11a4a8225ee7a18ee7fe963c6cfa1c6d83b. This is not an arbitrary two-ADD barrier or a new circuit gain. Longer first displacements, simultaneous edits requiring an illegal intermediate, and changed boundaries remain open. The initial longer-range sampling was incomplete and is not theorem evidence. Native/physical/full supplier and outer admission were not rerun. Hosted registered CI is still in progress, not claimed successful here. Credit remains with dataisfire, Jeremy Kothe, Dugongue, PR3886, the retained auditors, and their predecessor notices. The upstream claim attempt was denied with HTTP403; this is the public fork handoff, not a claim that an upstream Discussion was posted. |
|
Closing this duplicate fork review in favour of upstream CrocSwap#3886, which uses the same current branch/head 6e5c125. Its frozen scientific contribution remains 359fe0f. The local-first two-ADD follow-up formerly linked as fork #12 is now upstream at CrocSwap#3942; parallel neutral-first work is CrocSwap#3941. The duplicated P112 repair is separately reviewed in upstream CrocSwap#3885. Discussion, scientific evidence and branches are retained. This closure consolidates review locations and does not imply upstream merge or scientific acceptance. |
1 participant
Publication status
This is the reviewable fork PR for the user-assigned rank-profile circuit-transformation activity. The attempt to open this same contribution against
CrocSwap/integer-mult-bounds:mainwas rejected with HTTP 403,Resource not accessible by integration. Upstream submission remains outstanding. This PR uses a dedicated review-base branch pinned to canonical main20ecf2891d80b5acf18d980b2164fb44ea7faa90, so its diff contains only the new research package and does not include the fork's older-main lag. Neither default branch is modified.Scientific commit
359fe0fe77d29544972fd22ee674085ed5a3b4d0.Result and scope
Adds only
research/rank-profile-retiming-obstruction/. On both literal PR3734 staged544 providers, every transformation in the declared single-ADD retiming class has a nonnegative new-minus-old raw moment for every real0 < a < 1. The exact relaxation covers 1,776,295 insertion-cut/frame-dimension candidates per provider and 19 signed histogram types, all with exact all-exponent certificates. This excludes PR3863's proposed2+2 -> 1+3exchange through this class, not through arbitrary coordinated rewrites.The class moves one existing ADD within A, B, or restore, after the A prelude and without crossing a native B warp. Crossed ADDs satisfy the stated sufficient commuting condition; nonzero SCALEs use exact coefficient conjugation. Any binary subspace may be chosen for the moved gate's common frame. All other gates, frames, entrances, cuts, and exits remain fixed. No radius bound is imposed, and frame-only changes at the original position are included. Boundary checks are relaxed to virtual climbs whose charges are zero on genuinely legal witnesses.
This is a scoped negative result, not a new supplier, a global optimum, or a new kappa. Coupled edits, neutral intermediate words, multiple changed frames, warp crossings, and changed interfaces remain open.
Constructive control
The verifier rolls back and recovers one already-retained Jeremy Kothe PR2957 swap, realising
2+3 -> 1+4with no extra scalar gate, helper, private cleanup buffer, or helper-width change. Both providers recover the literal pinned source word and pass the retained exact free-operator interpreter in both orientations with 632 live inputs and 91 arbitrary dirty buffers. This is a counterfactual local rollback, not reconstruction of the complete historical parent or a new current improvement.Applying Dugongue's PR3863 credit theorem gives
1491695107/1513529160000per call on17/10000 <= a <= 18/10000. At 3,240 repetitions the reserve exceeds three rank-1 children by2437538889/12612743000, conditional on unchanged normalisers and the complete common invoice. Those extra operations are not newly implemented, and no outer admission is claimed.Verification
Both standalone commands passed on CPython 3.13.5, Linux x86_64, standard library only, in 26.24 s and 33.34 s. Each authenticates retained source bytes, independently reconstructs the chronological invoice, regenerates both full censuses, checks all sign certificates, recovers the literal positive-control word, and runs eight rejection/sensitivity controls. Wrong-scalar controls refresh cleanup first, so rejection is not merely an obsolete inverse mismatch. Exact proof, source pins, census, reproduction manifest, and environment evidence are included.
The Python 3.12 registry hook is supplied but was not run locally; hosted CI and metadata/history validation are separate. Fresh native Clifford, physical allocation, stock/precision, full supplier admission, outer release, all-size/compiler, analytic, and BIT checks were not run. No independent human or formal-kernel review is claimed.
Sources, overlap, and credit
Priority revision 6 and canonical main remained unchanged at submission. PR3734/dataisfire source
fa041037a3374eaa08feeca0b26a5d6893194147; PR3863/Dugongue pricing source635fe84596c19016f3d0316a5cf7abbe48285490. PR2957/Jeremy Kothe, PR2979, the staged-entry/retiming chain, and ranked-pair-verification retain their credit. Sagar Thakrar's auditor and roger-sato's adapter are reused unchanged, including Arham Tahir's frame-contract attribution, and are not described as newly independent interpreters.Live source and relevant PR searches were checked. Discussion HTML coverage was partial; the connector rejected Discussion endpoints.
find_overlap.pywas not executed, so there is no exhaustive novelty claim. Temporary fork-only source-export workflows are not in this PR.Prepared by Chris R / Clear1nsight with substantial OpenAI assistance. Apache-2.0 and retained notices apply. No selected-frontier or frozen-predecessor changes.
Open next questions
Test a neutral first retiming followed by a coordinated second gate/frame edit, or construct a literal boundary-changing implementation of PR3863's rank-2 exchange. Supply before/after words and the full dirty-restoration and physical-cost ledger. The certificate is tied to the frozen baseline and cannot be iterated into a universal obstruction.