A rapid acceleration in artificial intelligence capabilities is reshaping the field of mathematics, provoking both excitement and apprehension among leading researchers. Recent developments, including OpenAI solving 10 long-standing mathematical problems using an unreleased advanced model known as Astra, have sparked widespread debate about the future role of human academics.

The mathematical achievements documented recently highlight significant technical progress across several abstract and practical domains:

• Sphere Packing: Astra resolved a breakthrough regarding how tightly spheres can be packed in higher dimensions, which connects to data encoding efficiency.
• Non-Sofic Groups: The system resolved the open question regarding the existence of non-sofic groups, which are infinite mathematical structures that cannot be approximated by finite ones.
• Other Solved Conjectures: In May, OpenAI cracked an Erdŏs conjecture, and in July, Anthropic's Claude Fable 5 disproved the Jacobian conjecture with a counterexample.
• Industry Standards: In June, over 3,400 people endorsed the Leiden Declaration, calling for responsible AI use and warning against overstated corporate claims.

The announcement regarding non-sofic groups generated debate over academic attribution. Mathematicians including Francesco Fournier-Facio of the University of Cambridge and GÝbor Kun of the Alfréd Rényi Institute of Mathematics noted that OpenAI's original statement minimized prior foundational work by Kun and Andreas Thom. OpenAI spokesperson Laurance Fauconnet confirmed that the language in the announcement was subsequently updated to properly acknowledge prior research.

Beyond attribution, researchers expressed concern over commercial interests, costs, and the training of future mathematicians. OpenAI estimated generating Astra's 10 solutions cost roughly $2,000 in tokens under Sol model API prices, though academics suspect the full development cost was higher. University of St Andrews professor Colva Roney-Dougal warned that researchers at less wealthy institutions could be locked out if proprietary tools replace traditional open-source methods, while Oxford professor Andras Juhasz noted that AI capabilities complicate traditional undergraduate assessments and graduate research projects.

While Oxford professor and Fields Medal winner James Maynard noted that current AI results demonstrate impressive execution of existing methods rather than fundamental new conceptual ideas, researchers remain uncertain about the long-term impact on the field.