OpenAI stated that a new internal model solved more than 100 open mathematical problems after one month of training. The available source provides no details on the specific problems, methods used, or verification of the results.
Amid criticism from mathematicians, the company supports an independent advisory group at the Institute for Advanced Study; however, the pace of OpenAI's own research falls outside the scope of issues this group is mandated to address.
The story is significant because the claimed result impacts assessments of AI capabilities in mathematics and frameworks for independent oversight, yet current data is insufficient to confirm the quality, novelty, or reproducibility of the solutions.
editorial commentary
Why it matters
If the claim is confirmed, the next significant signal will be the publication of the problems, proofs, and independent evaluation. Substantial uncertainty remains for now: only one retelling is available without primary data, and the advisory group's role does not cover the pace of OpenAI's research.