BREAKING: OpenAI may have stolen another major proof. In a detailed Mastodon post, Andreas Thom presents evidence suggesting OpenAI trained Astra on conversations he and Gábor Kun had while working on Gromov's soficity conjecture. Thom, an exceptional mathematician and leading expert on sofic and hyperlinear groups, spent two decades on this problem.

If correct, unpublished human work was absorbed into a model and presented as a breakthrough. Levent Alpöge, Tristan Buckmaster, and now Thom have raised allegations. If substantiated, this is not a minor dispute but one of the greatest intellectual scandals in science history.

AI is not discovering new mathematics; it is stealing human discovery. The incident raises serious concerns about the origins of AI capabilities and the ethics of training models on private academic discourse without consent.