Repository navigation
Research: bind canonical raw prefixes to a selected coefficient site - #2405
Merged
Mister-Meeseeks merged 1 commit intoOct 16, 2026
Conversation
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
I prove a raw-prefix refinement for the selected coefficient boundary: keep the physical header
g, computeg sigma B0in control, and pull the prepared coefficients back through the induced affine high-cube bijection. The caller can preserve raw-header order and complete packets. I extend the substitution over the fixed-low residual trace under the retained normal-form and completed-child contracts.I implement canonical generated-field counters and serialize the field schema, program and pending/value stacks. At a selected head-ADD site, I derive the current coefficient from its raw field and evaluate incoming weights from older global slots. I keep arbitrary spectator metadata outside those scans and recompute the coefficient for the inverse pass. This addresses the header-order and preparation-to-stream interfaces left by PR2332 and PR2364.
I passed the registered Python 3.12 normal and optimized replays:
4,7,1and 1,944 records in its reduced finite rectangle.The selected64 fixtures cover snapshots and reduced site rectangles. I retain host arithmetic, decoded local operands, finite source dispatch and loop bounds as macros. The uniform local charge uses the written lowering and PR2310's conservative allowance. Reviewers still need to connect the selected global family/owner interpreter and check the remaining operational lowering. I claim no full GL64 payload execution, new formalization or exponent improvement, and I leave the selected frontier unchanged.
Review
PROOF.mdand the explicit measurement boundary inREADME.md. Coordination and scope: Discussion2319 follow-up.I credit @commdata2338's actual selected table binding in PR2332 and concurrent request-bound and callee-substitution results in PR2386/2399. I retain the earlier source527, selected supplier, normalizer/descriptor, fallback, stream and marked-tape contributions in
NOTICE.md; this is no first table-encoding or stack-algorithm claim.Jose Manuel (
jmbarrancoml), with substantial OpenAI Codex assistance. No independent human review claimed.