OpenAI Posts 722 AI-Written Math Manuscripts From Unreleased Model

The significance lies less in the manuscripts than in the disclosure gap: OpenAI is making a capability claim about a model it has not released, while the artifacts themselves are the only public evidence. That puts the burden on independent verification, and the immediate audience is the mathematics community, whose skepticism here echoes a recurring pattern in AI research announcements. Whether outside mathematicians can reproduce or validate the claimed solutions is the open question, and it is the one that will determine if this reads as a milestone or a marketing artifact.