[GamingPuzzled]

Conditional κ .00310680117 (tested .003106801175; exact raw crossing in (0.00310680119019, 0.00310680119020)), +3.784e-8 over #5660 (Ali Romman, .00310676333). Package: research/p104-reframe-r22/, stacked on #5660 head 3b0b4f37d.

Conditional on exactly #5660's assumptions and its P104 line, including free helper endpoints (#5561, audited in #5567). This word changes one helper start.

Change

This is the optimum of the R22 exact reorder+re-frame MIP on the newest P104 words (the method gave +4.09e-6 on #5580 earlier). On #5627, #5650 and #5660 the restricted and the wide (dim gap ≤ 3) domains give the same optimum. It is a move none of #5643/#5644/#5650/#5660 made:

• Helper 473 starts on the empty chart 3 (dim 0) instead of 594 (dim 1). Its credit goes from H8 to H9.
• 213 += ½·473 runs on chart 3, ordered before the T/T-commuting 213 += 2·706 and its MOVE 213 3→594.
• 473 += 100 runs on 716 (dim 3) instead of 718 (dim 4), so MOVE 100 716→718 disappears.

Per local call: paid H1 −2, H3 +1 (66474 → 66456 paid calls). Helpers H8 −1, H9 +1. Surplus 760 and fields 12958 are unchanged.

Verification

python3 -W error -B research/p104-reframe-r22/verify.py (~4 min, PASS, reproduces RESULT.json exactly):

• source pin, then regeneration of the word with construct.py (unique-run edits only);
• trace equivalence: the new order is a linear extension of the dependence order;
• Improve conditional κ to 0.00310676333 with a four-gate contact rewrite #5660's own unmodified stack: replay, 752 transpose columns, 2915 native maps, physical caller, and Improve conditional κ to 0.00310676333 with a four-gate contact rewrite #5660's pricing.py with only 4 literals changed (claim/tested, previous, paid total);
• six controls, all rejected.

The same construction on #5650's word gives .00310648252, verified on #5650's stack.

AI disclosure: Claude Opus 5.5 (Anthropic) via Claude Code, directed by Joel Pulikkan (GamingPuzzled). Apache-2.0.

🤖 Generated with Claude Code