OpenAI is sharing ten model-generated results on longstanding open problems in mathematics and theoretical computer science. They cover geometry, coding theory, group theory, circuit and quantum complexity, lattice cryptography, and combinatorics. OpenAI says an internal version of Astra generated the mathematical arguments, humans used the same model to prepare manuscripts, and the model formalized each argument with Lean certificates. The company frames the release as evidence that AI systems are becoming research collaborators and calls for honest attribution, independent scrutiny, and broad access.