OpenAI has announced that an advanced unreleased AI model, known as Astra, produced solutions to 10 long-standing mathematics problems, some of which had stumped academics for decades. The development has stirred a mixture of excitement and unease among mathematicians, including Oxford's Fields Medal-winning professor James Maynard, who describes grappling with what the shift means for a field traditionally slow to change. While many see huge potential for accelerating discovery, others fear for the future of human mathematicians and the next generation entering the discipline.

The problems tackled ranged from abstract theory to questions with real-world applications, including sphere packing in higher dimensions (relevant to data encoding), improvements to error-correcting codes used to recover information from noisy signals, and resolutions to questions about complex network structures. Other results touched on quantum game theory and high-dimensional search problems with implications for post-quantum cybersecurity. One notable result, on the existence of "non-sofic groups", sparked controversy after Cambridge mathematician Francesco Fournier-Facio and colleagues argued OpenAI's announcement understated the contribution of the human mathematicians whose prior work the AI had built upon.