AI Is Getting Good at Math. Here’s Why.
The essay reports OpenAI’s internal model disproving the Erdős unit-distance conjecture (open since 1946), proving the Cycle Double Cover conjecture, and solving ten more open problems — and analyzes why: models search depth-first with automated verifiers, trading time for compute, so ordinary hints become actionable and failed attempts become negative hints. Quoting OpenAI mathematician Jacob Tsimerman: AI can play for longer and in more treacherous waters than mathematicians without getting overwhelmed. It matters as a framework for predicting which problems AI will crack next.
SOURCES
- Latent Space2026-10-23 15:19