An unreleased Anthropic model made progress on one of math’s biggest unsolved problems

Anthropic's unreleased AI model made notable progress on the Riemann hypothesis by increasing the lower bound of solutions where it holds. The model autonomously coordinated 60 sub-agents, testing 650 ideas over 31 million steps, with key contributions from two sub-agents. The results were verified by Anthropic mathematicians and formalized using the Lean proof assistant.