Terence Tao Urges Math Overhaul by End-2026 as AI Solves 7 of 10 Unpublished Problems
Updated
Updated · New Scientist · Aug 6
Terence Tao Urges Math Overhaul by End-2026 as AI Solves 7 of 10 Unpublished Problems
3 articles · Updated · New Scientist · Aug 6
Summary
Months, not years: Terence Tao says mathematicians must reorganize their culture by the end of 2026 as AI rapidly erodes humans’ monopoly on solving hard problems.
Throughout 2026, AI has been cracking long-open questions at a pace of a few each week, creating what Tao calls a shift from proof scarcity to proof abundance and clogging peer review because validation remains scarce.
Tao argues the danger is not just more machine-generated proofs but a loss of meaning if AI both produces and formalizes arguments that computers verify yet humans cannot understand, teach or absorb into the field.
Education and professional norms are now at stake: Tao says students may need limits on AI use, warns tech firms are shaping mathematics’ public narrative, and notes 2026 Fields medalist Jacob Tsimerman is joining OpenAI.
Tao is pushing alternatives, including the First Proof project—where AI produced 7 publishable solutions to 10 unpublished problems—a new video-based journal, and the Leiden Declaration, which has drawn more than 3,000 signatories.
As AI floods mathematics with complex proofs, who will verify the truth when machines outpace human comprehension and peer review systems collapse?
If artificial intelligence can solve mathematics' greatest mysteries, will human mathematicians become obsolete or merely evolve into curators of machine logic?
From Proof Scarcity to Proof Surplus: The 2026 AI Revolution and the Future of Mathematical Research
Overview
In August 2026, OpenAI’s Astra model triggered a foundational crisis in mathematics by generating ten major breakthroughs, which were quickly reproduced by other AI systems. This leap was made possible by integrating large language models with formal proof systems, eliminating errors and enabling rapid proof generation. As AI began to automate routine and advanced proofs, the mathematical community shifted from proof scarcity to a surplus, overwhelming traditional peer review and leaving many results uninterpreted by humans. This forced mathematicians to focus on understanding and meaning-making, while AI handled technical tasks, highlighting the urgent need for new models of collaboration, validation, and education.