OpenAI, Anthropic, and a separate research team reported new AI-assisted formal mathematics/proof results using automated proof generation and verification
Research publication Provisional 64% confidence first seen
The coverage describes multiple recent outputs where AI systems generated and verified formal mathematical proofs, including long-running proof efforts and approaches that use repeated search with automated verifiers. It also highlights separate research that used AI agents to formalize statistical-learning-theory statements in Lean and generate substantial formal proof content, reporting compute and output scale.