AI Math Advances Shrouded in Secrecy, Scientists Urge Open Methods

rbhar90 · x · 2026-08-03

The author points out that the rapid advancement of AI in mathematics is currently accompanied by extreme secrecy. The public knows nothing about how models like Astra, Fable, and Mythos are trained, except that many mathematicians have been paid to generate training traces.

While this cloak-and-dagger approach benefits the IPOs of companies like OpenAI and Anthropic, it harms almost everyone else. The author calls on external scientists to strip back the secrecy and spread sunlight on these methods through open models, open datasets, and community-run organizations to understand the generality of AI reasoning and identify vulnerable scientific fields.

Original post →

More from AGI Musings

AGI Musings channel →