The published Bloch verification reproduces, and the author's own verifier accepts sqrt(3)/4 + 0.0153040536, so the published 0.0153 left 4.05e-6 of its own near-branch certificate unclaimed and that data has no further room.
hunts/bloch_ceiling (Record 61 of 98 in chronological sequence)
Guiding Question
How far above 0.0153 can the variable-radius dichotomy for Bloch's constant be pushed with the author's own cuts, grid and Arb verifier when ETA, LARGE_RAD and the point-certificate resolution are moved off their chosen values, and is the published verification sound as shipped?
Method & Verification
Archive bloch-computations-1.0.0.zip checksums were verified, the author's programs were reproduced locally and on Modal, and the unmodified certificate data were rerun at target 0.0153040536 across all 24 away sectors and 38,400 initial cells with zero refusals.
Lineage & Relationships
Primary Sources (at pin 8fa46e134)
Editorial Notes
The near branch is Arb-verified; the away floor and the moved-parameter ceiling near 0.015359 are INFERRED float search, so the composite grade stays measured. Sectors 0 and 1 at the published target match reference-run/logs to the integer (270,744 and 292,931). Cost 324 core-hours on GitHub Actions, billed zero. Nothing here bears on RH.
Date Provenance
commit 3d949739fa6ce1f4a89f8f7ebf827f537cdbb0fb, hunts/bloch_ceiling/RESULTS.md, author 2026-08-23T21:37:32-05:00