Part of the challenge with the AI math advances is the current air of secrecy around them. We don't know how Astra/Fable/Mythos etc are trai…

Part of the challenge with the AI math advances is the current air of secrecy around them. We don't know how Astra/Fable/Mythos etc are trained. We know many mathematicians have been paid to create training traces. How general is the reasoning? What fields (in mathematics and beyond) are vulnerable? The secrecy and cloak-and-dagger plays into OpenAI/Anthropic's hands for their IPOs but harms just about everyone else. The task for scientists on the outside is to strip back the secrecy and spread sunlight on these methods. Open models, open datasets, community run organizations are needed.

Source: Gary Marcus (X) | 2026-08-03