Math 2.0 Reframes AI-Era Mathematical Progress

Tao’s warning is less anti-AI than anti-scoreboard: proving problems faster does not automatically create mathematical understanding.

• –OpenAI’s public repository includes hundreds of AI-generated manuscripts, with Lean formalizations and plans for workshops to help mathematicians assess the results.
• –Developers building mathematical AI systems should optimize for readable proofs, citations, reproducibility, and follow-on discoveries—not just solved-problem counts.
• –Large-scale automated solving may “contaminate” open-problem ecosystems by removing alternative paths and discouraging researchers from pursuing questions that appear already solved.
• –The emerging opportunity is a human-AI research loop where models generate and verify conjectures while mathematicians supply judgment, context, exposition, and new questions.