AI models solve decades-old mathematical problems with minimal computational cost

Two major AI companies recently demonstrated advanced capabilities in mathematical research. Anthropic deployed Claude to identify cryptographic weaknesses, spending approximately $100,000 on token usage while conducting rigorous security research. OpenAI applied an internal version of its next-generation model to solve ten mathematical problems that had remained without significant progress for at least a decade. Each solution cost less than $2,000 in GPT-5.6 token prices. The results were disclosed with considerable transparency: the openai/ten-proofs repository contains formalizations of the findings in Lean 4, accompanied by a research paper and a PDF document in which the model reconstructs how each proof was developed based on its internal reasoning traces. Mathematician Terence Tao described this trend as the emergence of "big mathematics," where decentralized collaborations between humans and machines will transform the discipline. In this scenario, machines handle the technical heavy lifting while humans preserve the creative components.