OpenAI's 10 Math Breakthroughs Trigger Field Crisis
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI published '10 Advances in Mathematics and Theoretical Computer Science' alongside several hundred pages of documentation, claiming its new internal model Astra solved problems including quantum game theory and sphere packing in higher dimensions.
- Robert Hart reported the math community's reaction as a 'bombshell,' with researchers telling him that if a human had solved all 10 problems, 'we probably wouldn't believe it' — and these were problems 'mathematicians actually care about,' unlike previous AI breakthroughs.
- Astra, OpenAI's unnamed internal model, likely disproved the 80-year-old unit distance conjecture earlier in 2025, which Hart identifies as the moment the math field's anxiety escalated.
- AI models remain 'truly, truly terrible' at basic arithmetic — counting R's in 'strawberry,' telling time, identifying the day of the week — while producing work 'on par with good mathematicians' on abstract theoretical problems, a disconnect Hart calls the field's 'existential crisis.'
- OpenAI's blog initially claimed the 10 problems saw 'no progress in the last 10 years,' but the accompanying papers themselves credit prior progress from specific researchers — a framing discrepancy Hart flagged as raising attribution concerns.
Why it matters: Mathematicians told Hart that if a human had solved all 10 of OpenAI's claimed breakthroughs, 'we probably wouldn't believe it' — and the crisis extends beyond academia, since Hart explicitly raises whether these capabilities will transfer to other domains and whether academic funding structures for training new human mathematicians remain defensible if frontier models keep solving the field's hardest problems.
Ask SkimNews
