Repository navigation
Conditional κ = 0.0017193776586975866590: dirt-entry rewrite of A-phase helpers 460/462/478/480 on #2833 - #2853
Merged
Conversation
…30 and 542 on CrocSwap#2802's construction
…-phase helpers 460/462/478/480 on CrocSwap#2833 Extends CrocSwap#2833's (jeremykothe2) dirt-entry rewrite from phase B to phase A: four helpers (one per quartet) enter dirty at a 1-dim frame, their compensation is deleted and width drops 9 -> 8. New rebuild_borrow stage dirtentry4 with tamper controls, CP-SAT owner fill, strict assembly and regenerated outer ledger. +7.20222e-7 over CrocSwap#2833, +7.19381e-7 over CrocSwap#2843. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Contributor
|
one day we will flip the bind trust the process |
Contributor
|
I have teams of well trained electrons working for me night and day to solve this problem. |
Mister-Meeseeks
pushed a commit
that referenced
this pull request
Oct 16, 2026
…no kappa) Bounded screen of single-helper dirty entry, with every existing gate fixed, on the #2863 (PR2857 shared-kernel) and #2894 (preload-move) supplier words. Exact adjoint responses, full-program frame feasibility, partner cancellation, sub-frame entry, insertions in either phase and one-hop relays with an exact bilinear validity test. Result: 0 frame-feasible candidates on all four words; 8 apparent relays, 0 valid; single-gate and cascade frame refits 0 improving. Calibrated: the same screen gives exactly #2853's 16 quartet helpers on the PR2833 words. With frames relaxed, all 184 eligible helpers are algebraically cancellable, so frame placement is the obstruction; #2894's preload re-synthesis of two zero-frame-first quartets is consistent with this. Also records an independent fast and full cold replay of #2863. Vendored inputs keep their PR2833, PR2857 and PR2894 notices. Prepared for digitalis-p with substantial Claude (Anthropic) assistance; no independent human review claimed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Mister-Meeseeks
added a commit
that referenced
this pull request
Oct 16, 2026
Original authors and exact research bytes preserved. The only resolution combines reviewed additive research-package attributes; original attribute files are retained in the audit. Source-only intake and all remaining scientific obligations recorded in research/community-efficient2300-audit. No selected frontier change. PRs: #2948, #2856, #2853, #2312, #2302, #2288, #2281, #2269, #2265, #2220, #2184, #1677, #1672, #1661, #1640, #1601, #1586, #1536, #1513, #1510, #1479, #1458, #1444, #1377, #1274, #1261, #1172, #2950
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
5 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.
Conditional κ = 0.0017193776586975866590…, which is +7.20222e-7 over #2833 and +7.19381e-7 over #2843. It is stacked on #2833's head
e9307b48dand updatesresearch/gmx-t74-borrow43-handoffin place, following the same convention as #2843. The exact value (3046 decimals) iskappainouter/OUTER-CLAIM.json. Both gains are exact gates inverify_release.py.Credit first: #2833 (@jeremykothe2) is the base package and the source of the dirt-entry mechanism. It builds on #2802 (@mdumitrean), #2751 (@tekuuu) and #2766 (@DavidAriasC). Also credited: #2843 (@mdumitrean) and #2827/#2824 (@Dugongue). The full lineage is kept in the README.
Changed hypothesis. #2833's rewrite needed a helper whose first event is in phase B. This PR shows that a helper first touched in phase A can also enter dirty. The condition: its dirt response, minus its compensation, lies in the span of unit reads
z += ±k. These reads are inserted while k is still pure, at its entrance line or first frame, and stay frame-monotone for z. A-phase targets are excluded as z.new/borrow/dirtentry4/search/): 74 have their first gate at the zero frame, 36 carry dirt through warps, 63 fail the span test, and 16 pass. The 16 fall into four quartets. Each helper needs its own quartet as receivers, so exactly one per quartet composes: 460, 462, 478 and 480.What changed
rebuild_borrowstage,dirtentry4, which checks the exact gate edit: 12 inserted A-phase reads before each helper's first event, no change to B or the warps, exactly 28 compensation gates removed, the new entrance frames, and all other fields unchanged.DIRTENTRY4-CONTROLS), each rejected for its own structural reason.reconstruct.py.new/SPEC.json, then strict admission..gitattributesfiles are included and pinned), CLAIM and RELEASE.Replay (Python 3.12 with numpy, gmpy2, sympy 1.14, mpmath and psutil):
I ran this myself from a fresh work directory with no native cache, on a separate checkout of this branch. It ended with
PASS_RELEASE 0.0017193776586975866590…(about 15 min).Not composed: #2843 (T124) and #2848 (T248) only change the allocation layer, and their allocations are derived for #2833's widths. Composing either one means re-deriving it for the new inventory, which is worth about 1e-9. Left for a follow-up.
🤖 Generated with Claude Code