Mathematician pushes back on OpenAI PDE proof criticism: messy but salvageable

RexDouglass · x · 2026-09-13

Responding to criticism that OpenAI's proof about blow-up solutions of PDEs is bad, mathematician Scott Armstrong argues a reasonably understandable proof with an interesting narrative can likely be extracted from it, and that its Lean repo is shorter than the one he wrote for his last paper. He notes human-written papers in this subfield are themselves notoriously messy, so the verdict may not look bad once the dust settles.

Original post →

More from Research

Research channel →