Independent audit requested: Bernstein obstacle Mosco and free-boundary theorem

Corrected audit target

Please independently review the corrected Bernstein–Bézier obstacle theorem package at:

• immutable branch: review/bernstein-obstacle-v2-corrected
• commit: f2bd41f19ff5afbcca8a23f9afdcdf084364dae4
• correction PR: #106

The original PR #95 / v1 commit remains frozen for provenance, but new analytical reviews must use the corrected v2 target.

The complete adversarial checklist is in:

math/bernstein_obstacle/INDEPENDENT_AUDIT_REQUEST.md

The two principal claims are:

1. Mosco convergence of the coefficient-feasible Bernstein finite-element cones to H_0^1(Omega) ∩ {v >= 0}, with strong convergence of symmetric coercive obstacle minimizers.
2. Under the corrected explicit local grading, broken regularity, multiplier, and boundary hypotheses,
   ||u-u_h^B||_{H1} <= C (h^r + h_Gamma^(3/2)).

Material corrections reviewers must inspect

• R_h = {T : dist(T,Gamma) <= kappa h_T} uses local element size;
• the fixed one-ring patch is locally quasi-uniform and satisfies |omega_h| <= C h_Gamma;
• the bulk estimate assumes a uniform broken H^{r+1} bound outside the patch;
• physical-boundary coefficients are split into exact zero face coefficients and off-face coefficients controlled by inward linear growth;
• the phase-sensitive selected-mesh Hertz factor is no longer presented as a general claim.

Required reviewer output

• exact theorem and hypotheses accepted or rejected;
• pass/fail on each audit item;
• explicit references or derivations for every imported theorem;
• files and scripts executed;
• any counterexample, missing assumption, or corrected statement;
• separate assessment of mathematical validity and novelty.

Proposed bounty

• $7,500 for a complete written mathematical audit;
• $15,000 for the audit plus independent code reproduction and either a signed theorem endorsement or a concrete counterexample/fix for every failed item.

These are proposed project bounties, not standardized market-rate claims.