AI on open problems: verified results

Claims about AI solving open problems are easy to make and hard to check. This page lists only results that passed a check here — the Lean kernel or a certificate checker, an independent re-run, or a review that had to give reasons — with the model the contributor declared, how the result was checked, and how the person and the AI worked together. Negative results and literature finds count too.

Models are as declared by contributors. Download everything as CSV or JSON (problem status) or use the API. See also the external record of AI contributions to Erdős problems.