The upper-bound half reproduces as a primal rational C <= 9990167/26214400 = 0.381094627, 1.7e-4 above Haugland's 0.380926, with the m=32 stall being the solver rather than the family, while the convex-program half was not reproduced and the plain averaging skeleton caps four weights at 0.2526 or below.
hunts/r_828c8b (Record 65 of 98 in chronological sequence)
Guiding Question
is the published minimum-overlap record the limit of its own parameterisation, or the point at which the search stopped
Method & Verification
The m-piece minimax was run at m = 4/8/16/32 and the accepted object was re-evaluated in exact rationals, while the lower side was probed with uniform and Fourier averaging skeletons and four explicit weight-profile witnesses.
Lineage & Relationships
Primary Sources (at pin 8fa46e134)
Editorial Notes
Handback status is not-settled; RESULTS status is partly settled. Disposition follows RESULTS. Do not read front B float local minima as optima. White's actual program was not read here; that is overlap_lower. A shift-enumeration guard caught 2 of 3 planted faults, with the third recorded as a known miss.
Date Provenance
commit 241939a0e7ceef4a80570f41b1a87eb696236063, hunts/r_828c8b/RESULTS.md, author 2026-08-23T17:11:56-05:00