Kate Wenqi Zhu is a Postdoctoral Researcher at the Mathematical Institute, University of Oxford, who will give a talk on the topic of ‘AI for Mathematics: open problems, a workflow, and what it leaves behind’.
2026 has felt like a grokking moment for AI and mathematics. This talk is a working report from a numerical analyst, not an AI evangelist. I will describe (1) seven open problems from the numerical linear algebra catalogue that we resolved with AI agents, and how the community has responded; (2) the agent-to-agent workflow behind them, designed so that every proof is interpretable, traceable to its sources, and teachable, with independent cross-model review before anything is recorded as a result; (3) two worked examples: a deterministic eigenvalue-separation result whose proof pieces together convex analysis and classical numerical linear algebra, and a sum-of-squares result whose key idea we distilled into a reusable 'skill' for later problems; and (4) what I think this means for mathematicians: which problems suit AI, which suit people, and why asking the right question matters more than ever. I would like to leave time for an open discussion.