Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
54 changes: 54 additions & 0 deletions research/t37-integrality-audit/CLAIM.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,54 @@
# Local sprint claim

- Track: adversarial mathematical audit; OS assignment seed **65**.
- Base: `407d52ff15b2217b6e22bce44728b4c3907b6515` (origin/main fetched
2026-10-13). Branch: `research/audit-seed65`.
- Object: the integer-count/equality step in PR942's fixed-U8 all-replica
signed-dual relaxation, especially nonmultiples of 37.
- Method: derive the credit-slack lattice; certify a 37-state residue
potential with exact rational interval edge bounds; exhibit relaxation
witnesses and controls that remove one premise at a time.
- Deliverable: scoped proof, deterministic standard-library checker,
source pins, exact certificate and negative controls. No new kappa,
supplier, physical owner assignment or frontier edit.
- Expected finish: 2026-10-13 09:25 America/Los_Angeles. Reserve the final
quarter for verification, independent review and packaging.

The three candidates were (0) the prime-power cyclic inverse circle proof,
(1) the disjoint-helper matching floor, and (2) integer consequences of the
all-replica signed-dual support inequality. Seed `65 % 3 = 2` selects the
third. The inverse family already has several open audits; the matching
floor has a complete integer-dual review. The selected obligation has a
qualitative equality statement, but no quantitative residue certificate.

Curiosity Bot checked the issue listing and current PR metadata immediately
before this local claim. No pinned sprint coordination issue was visible;
the two open issues were #913 and #650. No issue was opened to reserve work.
Queries for integrality and all-replica work found no competing quantitative
residue audit. This is bounded coverage, not a claim to have read every PR.

Overlap: Jonathan Capalbo's #942 proves the all-T support bound and
constructs T37. Chafik Boukhalfa's #890 exhausts a finite T<=22 relaxation.
Joel Pulikkan's open #962 independently constructs T37 and identifies its
optimal normalized profile. This audit credits those results and adds the
slack lattice, a uniform sharp numerator penalty in the arithmetic
relaxation, and all 37 residue-class certificates. It does not repeat their
construction or claim priority for T37 optimality.

Delegated review is confined to this same integrality step and certificate.
Stopping rule: prove/check the finite 37-state certificate and declared
controls, then stop; no broader supplier or parameter search.

## Result and handoff

Completed 2026-10-13. The audit derives the 1200-unit slack lattice and
sharp off-37 numerator penalty, with all 37 residue certificates. Both
checkers passed normally and under `-I -O -B`: 1,163 bin patterns, 1,591
edge inequalities, 37 witnesses, and 8 primary / 10 independent rejection
controls. The fresh overlap check also distinguished PR974's interface
budget study. No physical construction or new kappa is claimed.

Submitted one focused research-audit PR:
[CrocSwap/integer-mult-bounds #992](https://github.com/CrocSwap/integer-mult-bounds/pull/992).
No coordination issue was available, so this is the single local claim
update with the result and link. The selected frontier remains unchanged.
Loading
Loading