Repository navigation
Certify local-first two-ADD retiming obstruction on staged544 - #12
Clear1nsight wants to merge 2 commits into
Conversation
…oviders Add exact interaction/new-gap reductions, 11,917,718 reduced prices per provider, full first-frame coverage, and source-bound rejection controls. No new kappa. Retain PR3886, dataisfire, Jeremy Kothe, Dugongue, and auditor lineage credit. Prepared by Chris R / Clear1nsight with substantial OpenAI assistance.
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.
|
Publication checkpoint, 2026-10-19 04:38 UTC. Current head is Current-head Registered research packages run39200158512 has passed the plan job, including onboarding tests and affected-package selection. The sole selected replay, |
|
Closing this fork-local review in favour of upstream CrocSwap#3942. The upstream PR contains only scientific commit 1416f8c and its 20 research files, with the CrocSwap#3941 overlap disclosed. The additional P112 workflow repair at this fork head is separately represented by upstream CrocSwap#3885. Research evidence, discussion and branches are retained. This closure removes a duplicate review location; it does not imply upstream merge or scientific acceptance. |
1 participant
Summary and scope
Follows the neutral-first two-ADD handoff from CrocSwap CrocSwap#3886 and fork #9. Adds only
research/staged544-local-two-add/, with a scoped proof, exact complete census receipts, source pins, and registered replay. No new circuit saving, supplier, or kappa.On each literal PR3734 staged544 provider, a first existing ADD may change its frame in place or cross one adjacent ADD/nonzero SCALE with any legal first frame. A second ADD may then move any distance inside its allowed scalar corridor and use any common binary frame. Every combined raw rank moment is non-improving for every real
0<a<1. The result covers neutral and locally losing first moves, not just neutral ones.The first displacement is limited to one neighbour. Both intermediate and final words must obey the fixed boundary contract. Every other gate/frame, entrance/exit, target/centre cut, and native warp remains fixed. ADD crossings use PR3886's sufficient commuting rule; SCALEs use exact coefficient conjugation. No warp or phase boundary is crossed. This does not exclude a longer first displacement, simultaneous changes requiring an illegal intermediate, added/deleted gates, moved SCALEs, or a changed consumer/interface. It is not an arbitrary two-ADD or global architectural barrier.
New contribution
The interaction reduction handles disjoint second operands and selecting the first ADD again. A new single-gap lemma shows that a fixed-frame adjacent swap can introduce a new price only for a third ADD inserted between the pair and touching both members. Changed first frames are enumerated as actual subspaces, because orientation matters to the second move. All second frame dimensions are covered without sampling bases.
Per provider, the exact first census contains 5,756 nontrivial first-move records: 4,709 fixed-frame adjacent swaps; 49 neutral and 616 losing adjacent changed-frame moves; and 29 neutral and 353 losing stationary frame changes. The 1,047 changed-frame first words require 11,903,025 interacting second cut/dimension checks. The fixed-frame reduction adds 14,693 new-gap cases. Total 11,917,718 reduced candidate prices per provider, with 49 signed histogram types, all exactly certified for the full real interval. Counts are price cases, not distinct complete circuits.
The primary verifier independently regenerates each provider rather than assuming their histograms or binary frames agree. It also regenerates all 1,776,295 PR3886 single-move cases as an explicit premise. First-frame quotient dimension is at most three; a resource bound raises an error rather than silently truncating.
Verification
Both full standalone censuses passed on CPython 3.13.5, Linux x86_64, standard library only, in 344.93 and 348.66 seconds. Their complete external receipts were retained byte-for-byte. The source words pass the retained exact free-operator interpreter in both orientations on 632 live inputs plus 91 arbitrary dirty buffers. Each changed-frame intermediate invoice agrees with the separately retained frame auditor.
Normal and assertion-disabled controls both pass. They include independent quotient-basis/subset-XOR enumeration of all 3,118 first-frame intervals; 26 fast-bridge/full-trace comparisons including every nonzero bridge case; 3,804 unpruned non-bridge prices; 48 exact scale-conjugation matrix checks; literal undo, source corruption, frame, and scalar-barrier controls; and actual production-predicate disable sensitivity.
The already-retained Jeremy Kothe PR2957 rollback/recovery control still realises
2+3 -> 1+4, recovers the literal source word, and is rejected by a non-improvement certificate. It is not a new current improvement. Wrong-scalar controls rederive cleanup first, so rejection is not caused by a stale inverse.The pinned registry schema validation passed locally with Git-pin checking disabled, and the original task card separately matches the authenticated base catalogue. Full registered Python3.12 execution and Git-history validation are delegated to hosted CI and are not claimed from that local check. Published code, proof, results, and metadata blob identities match the tested local package. Initial incomplete longer-range exploration is explicitly disclosed and is not used as evidence for the theorem.
Cost and remaining obligations
This is a fully reconstructed raw rank comparison under unchanged normalisers and common costs. Added nonnegative charges cannot turn it into a saving. Helper widths, buffer identities, and boundaries are fixed by the class. This does not admit every edited word as a complete physical supplier. Fresh native Clifford phases, physical allocation, uniform coefficients/stock, precision, all-size compiler, complete supplier/outer release, and analytic/BIT chains were not run. No independent human or formal-kernel review is claimed. The selected frontier is unchanged.
The next useful work must change a stated premise: a concrete first move across at least two neighbours, a simultaneous two-frame change, or a changed producer/consumer boundary with its complete dirty-restoration and physical bill. Repeating the exact local-first domain with additional seeds cannot change this result.
Sources, overlap, and credit
Scientific commit
1416f8cd6bfb2f2eb9aa5cc22f0388a85fdbac91, based directly on canonical main20ecf2891d80b5acf18d980b2164fb44ea7faa90, priority revision 6. The fork review base is pinned to that same canonical main, so the PR excludes the fork default branch's lag. No default branch or frozen predecessor package is changed.Sources are pinned to dataisfire/PR3734
fa041037a3374eaa08feeca0b26a5d6893194147and PR3886359fe0fe77d29544972fd22ee674085ed5a3b4d0. Retain Jeremy Kothe/PR2957, Dugongue/PR3863, PR2979, and the staged-entry/retiming lineage. Sagar Thakrar's auditor and roger-sato's adapter are reused unchanged, including Arham Tahir's frame-contract attribution. The trace, single-move theorem, and sign criterion are credited predecessor work, not newly independent implementations.The official overlap helper refreshed 630 open PRs and 2,544 Discussion records with complete pagination at 03:39 UTC on 19 October 2026. Local-note coverage was limited to exported sources. Canonical main and updated retiming PRs were rechecked before submission. No exhaustive historical/diff novelty claim is made. An upstream claim comment was denied with HTTP403; no successful claim post is asserted. Temporary fork-only source-export infrastructure is excluded.
Prepared by Chris R / Clear1nsight with substantial OpenAI assistance. Apache-2.0 and all retained notices apply. This fork PR is prepared for the user to submit upstream.