OpenAI and Anthropic achieve significant mathematical breakthroughs using AI models

Internal models at OpenAI found a counterexample to Erdős's unit distance conjecture in May, followed by a proof of the Cycle Double Cover (CDC) problem and solutions to ten other open problems by July, and a proof of Navier Stokes in September. Anthropic also published a formalization of Fermat's last theorem and advanced bounds on Riemann's hypothesis. Experts suggest AI excels at problems requiring a limited sequence of insights, such as completing derivations within an established conceptual framework, rather than those needing broad intuition like General Relativity.