Repository navigation
Finite bit construction: multi-write shear programs for all ten R439 phase-B blocks (24,305 -> 24,053 ADDs, no new kappa) - #1015
Merged
Mister-Meeseeks merged 2 commits intoOct 14, 2026
Conversation
…5 -> 24,053 ADDs, no new kappa) One deterministic three-phase search (min-excess greedy, random legal trajectories, beam) over the full multi-write shear class on the pinned PR589 tuned word replaces every phase-B block with a shorter exact program on the same registers and frames. Independent verifier re-derives each integer block map and Boolean truth table, splices all ten blocks with literal reverse cleanup, and rechecks 661 dirty-input endpoint columns, H1-H5, the paid profile (rank mass 5,129 unchanged) and the exporter bills: ADDs 24,305 -> 24,053, forward bound 49,347 -> 26,653, reverse bound 1,113,070 -> 776,006, payload 93 -> 90 bits. Eight mutation controls rejected. No minimality is claimed; frame 520 does not beat open PR CrocSwap#923. Prepared with help from opencode.
Contributor
|
100-PR batch review at Merged to Reviewed scope: Ten same-frame P111 phase-B replacements with exact integer transfers, all dirty inputs and literal cleanup; local cost only. The pinned scoped command passed; the batch ledger contains its command, input hashes, environment and log. This incorporates the research/tooling contribution and does not change the selected κ or discharge the inherited multiplication-theorem assumptions. Batch review, attribution and evidence. Thank you for the contribution. |
Mister-Meeseeks
pushed a commit
that referenced
this pull request
Oct 14, 2026
…ks retained via #1062 (no new kappa) Main retains open #1015's CANDIDATE.json verbatim (blob 16532f2) through merged #1062, where its nine phase-B block programs supply 244 of #1062's 283 saved ADDs, but #1015 had no ledger row, CONTRIBUTORS credit or retained NOTICE. Add a hash-checked receipt, retain #1015's NOTICE.md byte for byte, and add one CONTRIBUTORS credit line. No ledger, disposition or construction change; #1015 stays open.
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.
Status: finite bit-circuit construction; no new kappa. One deterministic search in the full multi-write shear class replaces every phase-B block of the pinned PR 589 tuned R439 word with a shorter exact program on the same registers at the same frame. The complete exported local word falls 24,305 -> 24,053 ADDs (388 -> 262 block gates, 252 complete ADDs with literal reverse cleanup).
Per-block (word -> this search -> best previously recorded): 506 26->21 (#891 class min 24; PR 1001 also 21), 507 35->24 (PR 880 30; PR 884's README's unpublished 24-gate hint matched), 519 42->27 (PR 908 32), 508 37->21 (#873/#891 30), 520 41->37 (PR 923 claims 29 - not beaten here), 509 31->21 (PR 973 also 21), 521 47->30 (PR 884 31), 522 43->29 (PR 906 33), 523 45->26 (#891 39, PR 857 40), 524 41->26 (PR 914 27).
Class and coverage. A block replacement is exactly a legal nonnegative row/column subtraction path from the block's integer transfer matrix to the identity; the class is all multi-write shear programs on the block's own registers (shared intermediates included), strictly containing #891's single-write class and PR 880's fixed splits.
search.pyruns per block: 6,000 min-excess greedy reductions with random tie-breaks, 4,000 uniform random legal trajectories (cap 90), one width-6,000 beam below the incumbent. Seeds20261013+frame; all trials complete; no adaptive stopping. Best found is not a minimality result for any block.Reproduce (stdlib Python 3.11+, ~2 min):
Expected:
PASS_R439_BSECTION_MULTWRITE gates=388->262 ADDs=24305->24053 columns=661 controls=8 rank_mass=5129 kappa=unchanged.search.pyregeneratesCANDIDATE.jsonbyte for byte (SHA-256771F980643402EB44B25312923B376A952E9C8ABB55796CBDC348BA51FC62728, confirmed across two full runs). The verifier imports no search code: source pins, per-block integer maps and 2^10-2^12 Boolean truth tables, ten-way splice with literal cleanup, all 661 dirty-input endpoint columns, H1-H5 via the pinnedxcheck_p.py, paid profile/rank mass, exporter bills and bounds via the pinnedexport_records.py, and eight rejected mutation controls. Exact report pinned inCERTIFICATE.json.Overlap (refreshed at submission). PR 1001 (506@21) and PR 973 (509@21) independently reach two of these counts - convergent evidence, credited. PR 989 composes four blocks (507/508/519/524) at 24,233 ADDs; this search's programs on those frames (24/21/27/26) are strictly shorter than what it composes (30/30/32/27) and this word covers all ten blocks. PR 923's frame-520 program (29) is stronger than this search's 37 and is not imported. The phase-A prefix line (PR 904/PR 936/PR 952/PR 980) and the D-prefix substitution candidates (PR 915/PR 926/PR 941) are different mechanisms, not composed. No concurrent implementation is used.
Scope. Only
research/r439-bsection-multwrite/is added;upstream/, the selected record and all consumed sources are unchanged (base007664306fff84fbb1c9a1b4053b69f439764034, pins inSOURCES.json). Conditional finite construction: native 5/9/11-stage replays, hosted invoice, supplier binding, outer assembly, Lean and remote CI are not run here; all inherited all-size/compiler, common-label/mutable-source, whole-call hosting, precision/tape and analytic assumptions remain. The spliced word is a research witness, not a proposed frontier. Sprint claim, method, seed (16, bit-supplier track) and development defects are recorded inCLAIM.md/VALIDATION.md.Credit: PR 555 / PR 543 by eumemic, PR 589 / PR 545 by DreamingOfClouds, Dylan Crow, Zhihao Chen and the retained supplier lineage; block-level framing and comparison credit to Aaron Wang (#891) and PR 973 by Cinna Davis (1337isnot1337, with OpenAI Codex assistance), PR 880 by Krishnendu Purohit (Deftsalt-debug, with OpenAI Codex assistance), PR 884 by martin Rv (martinxxx9), and PR 857 by zeusinsight, #873, PR 906 by CubeJerry, PR 908 / PR 914 / PR 923 by Ross Richardson, PR 989 by woklerx, PR 1001 by ToikY. Prepared with help from opencode. Apache-2.0; no independent human review is claimed.