OpenAI on Tuesday released 722 math manuscripts generated by its unreleased AI model on GitHub, claiming that most of them were produced with a single prompt. However, some academics, including MIT mathematician Andrew Sutherland, have expressed skepticism, stating that the claims remain unverified until the model and results are reproducible, and have requested more validation from OpenAI. OpenAI acknowledged that not all manuscripts have been computer-verified, with 162 (approximately 22%) having undergone mechanical logic checks using Lean software.