Day Old

AI news · Saturday, August 1, 2026

OpenAI’s new Astra model cracks ten long-standing, complex mathematical problems

OpenAI just pulled back the curtain on Astra, its next generation of AI models, by having an internal version solve ten mathematical problems that have stumped researchers for over a decade. These aren't just parlor tricks; the model tackled high-dimensional geometry, group theory, and quantum complexity, with the entire compute cost for all ten solutions ringing in at roughly $2,000. While the company is framing this as a massive leap for scientific reasoning, it’s also trying to thread a needle on credit: it admits the model generated the core arguments while humans just formatted the papers and double-checked the proofs via Lean, a software that verifies math. OpenAI says it wants to be transparent rather than claiming human authorship for work that is fundamentally machine-driven. The math community, meanwhile, is having a very public existential crisis about whether this is the start of a 'golden age' or the beginning of the end for the human mathematician. Some researchers are leaning into it as a productivity boost, comparing the model to an indefatigable PhD student, while others like mathematician Kirwin Hampshire describe the feeling as 'the rug being pulled out' from under them.

Elsewhere, the 'hardcore' culture of the AI gold rush is hitting the physical limits of its own leaders. Lilian Weng, a cofounder of Thinking Machines Lab, is the latest high-profile executive to step down citing health, joining a list that includes OpenAI's Fidji Simo and xAI's Greg Yang. Experts say this is a classic Silicon Valley catch-22: the industry self-selects for an 'all-or-nothing' work ethic that is fundamentally incompatible with chronic stress, even as those same execs push AI as a tool for personal productivity. That obsession with speed also bled into the consumer world this week, as OpenAI CEO Sam Altman faced a wave of online eye-rolling for suggesting parents use ChatGPT to generate daily podcasts about their kids' lives. The backlash was swift, with animator Alex Hirsch capturing the collective mood by simply asking, 'What if you just talked to your children?' Finally, in the legal trenches, a Minnesota judge refused to block a state ban on 'nudify' apps, ruling that xAI’s attempt to halt the law days before it took effect was too little, too late to prove 'immediate' harm. The case highlights the growing legal gray area of who is responsible when AI tools are used for harmful, non-consensual content.

The quick hits

Sources

Get the day's AI news in one calm read, every day.

Get the app on Google Play Get the app on the App Store Or read today's brief in your browser