A new development in mathematics has sparked a debate about the growing role of artificial intelligence in solving difficult problems. On August 31, mathematician Julia Stadlmann from the University of Illinois Urbana–Champaign made progress on the famous twin prime conjecture, a long-standing problem in number theory.
The twin prime conjecture suggests that there are infinitely many pairs of prime numbers that are exactly two numbers apart. Examples include (3, 5), (5, 7), and (17, 19). Although mathematicians have studied the problem for many years, proving that there are infinitely many such pairs remains an unsolved challenge.
Stadlmann’s work was considered an important achievement after more than a decade without major progress in this particular line of research. Mathematician Kannan Soundararajan praised the difficulty of the work and highlighted Stadlmann’s persistence in continuing to study the problem.
However, the situation changed very quickly. Within just three days, AI startup Axiom Math announced results that improved on Stadlmann’s findings. The pace became even faster when OpenAI reportedly produced a result that went beyond both earlier records only two hours after Axiom’s announcement.
The speed of these developments has raised concerns among some mathematicians. They argue that mathematics is not simply about setting new records or producing better numerical results. For many researchers, understanding why a result works and developing new ideas are just as important as reaching a stronger result.
Some mathematicians feel that AI companies may be approaching mathematical research differently from traditional academic researchers. Instead of spending years developing an idea and building a deep understanding of a problem, AI systems can rapidly search through large numbers of possibilities and produce results.
Renowned mathematician Terence Tao has also expressed concerns about this approach. He criticized the focus on quickly surpassing human achievements, arguing that mathematical research is also about understanding, curiosity and expanding human knowledge.
The episode has highlighted the changing relationship between humans and AI in mathematics. Human mathematicians often spend years working on difficult problems, developing techniques and learning from failed attempts. AI systems, meanwhile, can examine mathematical possibilities at extremely high speed.
Supporters of AI in mathematics believe these tools could help researchers solve problems that have remained unanswered for decades. They argue that AI can act as a powerful assistant, helping mathematicians discover patterns, test ideas and explore difficult calculations.
The debate is therefore not simply about whether AI is better or faster than humans. It also raises a deeper question about what should count as progress in mathematics. As AI becomes more capable, mathematicians will have to find ways to combine the speed of machines with the creativity, reasoning and understanding of human researchers.
The recent developments surrounding the twin prime conjecture show how quickly this field is changing. What once took mathematicians years of work may now be challenged by AI systems within hours, creating both exciting opportunities and difficult questions for the future of mathematical research.