From the #469 adversarial review: test_1072_free_surface_spherical.py is capability-only by design — its mean-free assertion is true by construction (_surface_mean of a _demeaned field using the same weights) and the |shape| <= 1.5|h_inf| bound survives a sign flip, so CI cannot catch a 3D sign error or a rate regression.
Reference numbers measured in the review (shell R_i/R_o = 0.547, cell size 0.35, radial gravity, Y20 initial topography): exponential decay with fitted rate 0.0842 vs half-space Cathles rate ρg/(2ηk), k = √6/R → 0.204; ratio 0.41 — an O(1) shell correction below the half-space rate, in the physically correct direction. A follow-up test should fit the decay rate over a few steps and assert the ratio lands in a band around 0.41 (resolution-matched), which pins both the sign and the shell physics.
Underworld development team with AI support from Claude Code
From the #469 adversarial review:
test_1072_free_surface_spherical.pyis capability-only by design — its mean-free assertion is true by construction (_surface_meanof a_demeaned field using the same weights) and the|shape| <= 1.5|h_inf|bound survives a sign flip, so CI cannot catch a 3D sign error or a rate regression.Reference numbers measured in the review (shell R_i/R_o = 0.547, cell size 0.35, radial gravity, Y20 initial topography): exponential decay with fitted rate 0.0842 vs half-space Cathles rate ρg/(2ηk), k = √6/R → 0.204; ratio 0.41 — an O(1) shell correction below the half-space rate, in the physically correct direction. A follow-up test should fit the decay rate over a few steps and assert the ratio lands in a band around 0.41 (resolution-matched), which pins both the sign and the shell physics.
Underworld development team with AI support from Claude Code