External reviewers wanted: signed Bernstein obstacle audit reports

Purpose

Recruit genuinely independent reviewers for the final Bernstein–Bézier obstacle theorem package using math/bernstein_obstacle/EXTERNAL_REVIEW_PROTOCOL.md. No email campaign is being used.

Final frozen review target

• immutable branch: review/bernstein-obstacle-v4-final
• commit: 61594952ad880d2b61759cfa93a19df979183c09
• analytical integration PR: #114

Reviewers beginning now must audit V4, not the moving research branch or superseded v1/v2/v3 statements.

Material V4 improvements

1. Corrected local-size risky set and one-ring geometry.
2. Uniform broken H^{r+1} assumption.
3. Shared-DOF support enlargement from R_h to omega_h.
4. Multi-face physical-boundary proof covering corner elements.
5. Tubular, compact-interior, and physical-boundary positivity split.
6. Explicit multiplier sign convention and exact energy identity.
7. Constructive scheduled-stage Lean theorem from mesh thresholds.
8. ThresholdSobolevFEMRecoveryData connecting local analytical estimates to the verified Mosco/minimizer endgame.
9. Corrected Hertz phase-sensitivity data and project-internal replication wording.

Candidate PR #117 ports the coordinate-free Hilbert VI and nested recovery-to-strong-convergence theorem onto V4. Reviewers should treat it as a candidate addition until its pinned audit is green.

Review panels

• #98 — prior-art collision — FUN BUX: 3,500
• #99 — Mosco/FEM recovery — FUN BUX: 4,500
• #100 — free-boundary clipping and h_Gamma^(3/2) — FUN BUX: 4,500
• #101 — Lean statement faithfulness — FUN BUX: 2,500

These are unfunded internal effort/complexity labels. They are not real or promised bounties. No payment is represented as available unless separate written funding terms are posted.

How to volunteer

Post your panel, expertise/profile, conflicts, and confirmation that you will review commit 61594952ad880d2b61759cfa93a19df979183c09.

A final report must state PASS, PASS AFTER CORRECTION, or FAIL, answer every panel item, and post the report artifact SHA-256. Counterexamples and corrected narrower theorems are fully eligible.

Independence boundary

Repository-owner reviews, automated-agent outputs, and project-internal implementations do not count as independent endorsement. The theorem will not be advertised as externally confirmed until qualified reviewers post public signed reports.

Issue #103 separately requests outside computational reproduction; the scikit-fem implementation in the repository is project-internal, not clean-room independent validation.