OpenAI model cracks 80-year geometry conjecture

This is a real research milestone if the verification holds up: not just a benchmark win, but a model producing a novel proof that experts accepted as credible.

• –The notable part is autonomy: OpenAI says the result came from a general-purpose reasoning model, not a math-specific system or proof search wrapper
• –It strengthens the case for reasoning models as research collaborators, especially in domains where outputs are formal enough to verify
• –The claim is still narrower than the headline suggests: it solves one precise geometry problem, not “math” in general
• –The practical signal for developers is that better test-time reasoning can surface unexpected structure, not just better pattern matching
• –Expect heavy scrutiny here; breakthroughs in theorem-land only matter if they remain reproducible and formally checkable