AI Co-Mathematician: Accelerating Mathematicians with Agentic AI
2026-05-07 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors created the AI co-mathematician, a tool that helps mathematicians work with AI to explore and solve math problems step-by-step. It supports tasks like coming up with ideas, searching for related work, running calculations, proving theorems, and developing new theories. The system keeps track of what works and what doesn’t while producing math outputs that feel natural to researchers. In early testing, it helped solve tough problems and find new research paths. The authors also showed that it performs very well on challenging math problem tests compared to other AI systems.
AI co-mathematicianinteractive mathematical workflowstheorem provingcomputational explorationresearch ideationstateful workspacemathematical artifactsproblem-solving benchmarksFrontierMath Tier 4
Authors
Daniel Zheng, Ingrid von Glehn, Yori Zwols, Iuliya Beloshapka, Lars Buesing, Daniel M. Roy, Martin Wattenberg, Bogdan Georgiev, Tatiana Schmidt, Andrew Cowie, Fernanda Viegas, Dimitri Kanevsky, Vineet Kahlon, Hartmut Maennel, Sophia Alj, George Holland, Alex Davies, Pushmeet Kohli
Abstract
We introduce the AI co-mathematician, a workbench for mathematicians to interactively leverage AI agents to pursue open-ended research. The AI co-mathematician is optimized to provide holistic support for the exploratory and iterative reality of mathematical workflows, including ideation, literature search, computational exploration, theorem proving and theory building. By providing an asynchronous, stateful workspace that manages uncertainty, refines user intent, tracks failed hypotheses, and outputs native mathematical artifacts, the system mirrors human collaborative workflows. In early tests, the AI co-mathematician helped researchers solve open problems, identify new research directions, and uncover overlooked literature references. Besides demonstrating a highly interactive paradigm for AI-assisted mathematical discovery, the AI co-mathematician also achieves state of the art results on hard problem-solving benchmarks, including scoring 48% on FrontierMath Tier 4, a new high score among all AI systems evaluated.