Artificial intelligence is transforming mathematical research at an accelerating pace, moving beyond calculation to genuine discovery. Mathematicians worldwide are now collaborating with AI systems that can propose novel approaches, generate working code, and uncover patterns invisible to human intuition alone. This partnership represents not just a new tool, but a fundamental shift in how mathematical knowledge might be created.
The Unequal Access Challenge
As these powerful systems emerge, concerns about equitable access are growing. José Simental of the National Autonomous University of Mexico highlighted this issue, noting that his opportunity to work with AlphaEvolve came only through a semester spent in the United States. "I think we as a community have to be very careful on ensuring equitable access to these sort of tools," he emphasized. This disparity raises important questions about who gets to participate in the next wave of mathematical discovery and whether AI might widen existing gaps between well-resourced and developing mathematical communities.
An Unexpected Discovery
In a striking example of AI's potential, mathematicians Jordan Ellenberg and Geordie Williamson collaborated with DeepMind researcher David Wagner in late 2025. They asked AlphaEvolve—a system not publicly available—to analyze Bruhat intervals within permutation groups, mathematical structures related to symmetry and arrangement. The researchers expected the AI to help with calculations, but what happened next surprised them.
After running overnight, the system had produced something remarkable. "In the morning, we were like, this program is really doing something interesting," Williamson recalled. The team exchanged a flurry of emails as they tried to understand what the AI had created. Rather than simply crunching numbers, AlphaEvolve had generated approximately 50 lines of Python code in its attempt to find intervals with specific mathematical properties known as large d-invariants.
AI's Internal Dialogue
Perhaps most intriguing was how the system approached the problem. AlphaEvolve engaged in what appeared to be self-directed reasoning, commenting internally about its process. At one point, it mused about proposing "something truly outlandish," describing it as a "'Crazy Ivan' maneuver for this problem"—a reference to a sharp submarine turn popularized in Tom Clancy's The Hunt for Red October. This metaphorical thinking suggested the AI was approaching the mathematical challenge with creative, almost narrative-driven strategies rather than purely algorithmic ones.
As the mathematicians examined the generated code, Ellenberg noticed something peculiar: when the number of elements in the permutation group was a power of two (like 16, which is 24), the program became dramatically simpler—shrinking to about five lines. "You can analyze it very explicitly," Williamson noted. "It's doing something very beautiful." The AI had apparently discovered an underlying mathematical elegance that only manifested under specific conditions, revealing a pattern human researchers might have missed.
Broader Implications for Mathematics
This discovery represents more than just a single mathematical insight. It demonstrates how AI systems can serve as genuine collaborators, proposing approaches human mathematicians might not consider. The ability to generate working code that implements novel mathematical ideas suggests AI could become a creative partner in exploration, not just a verification tool. This development comes alongside growing discussions about computer verification in mathematics and how technology is changing the field's fundamental practices.
The episode also raises fascinating questions about mathematical intuition. AlphaEvolve's "Crazy Ivan" comment suggests AI systems might develop their own conceptual frameworks for approaching problems, potentially leading to entirely new mathematical methodologies. As researchers continue to explore abstract mathematics finding real-world applications, AI-generated insights could accelerate connections between pure theory and practical problems.
The Future of Mathematical Collaboration
Most mathematicians agree this is just the beginning. As AI systems become more sophisticated, they may uncover patterns across diverse mathematical domains, suggesting connections between fields that have developed separately for decades. The challenge will be ensuring these tools enrich mathematical community rather than creating new barriers. The experience of researchers like Simental highlights the importance of making advanced AI tools accessible globally, not just at elite institutions.
What makes this development particularly significant is that AI isn't merely proving conjectures humans have already formulated—it's helping generate new mathematical objects and relationships. This represents a shift from computational assistance to genuine co-discovery, potentially accelerating progress in fields from number theory to topology. As with previous revolutions in mathematics, from Cantor's infinity to modern computer-assisted proofs, the integration of AI will likely transform not just what we can prove, but how we conceive of mathematical truth itself.
