The ability of AI systems to produce new mathematics is advancing at a breathtaking pace. Mathematics as a product of human culture and as a profession is undergoing profound changes. For the sake of clarity, all discussion below reflects my own personal views.
My own (evolving) thoughts include:
"One-shotting" results (prompting for a full solution to a problem, often without having the background to understand the problem or the potential solution) should be strongly discouraged, and I commit not to do it myself: human understanding is advanced through struggling through/having fun with a problem, and a checkmark that says "solved" has by itself limited value and can often be a net negative. If one-shotting becomes the norm, it is very hard to see how a shared human understanding of mathematics can survive.
OpenAI and Anthropic are having a profoundly destructive influence on the mathematics commons (partly but not only due to their one-shotting with models only available to them). Mathematicians should not collaborate with them in any shape or form.
I use AI to help with literature search, error checking, "technical" aspects of my work (e.g. implementing the technical details of a strategy my co-authors and I developed) and "idea brainstorming" in a similar way to what I would do with a colleague. I acknowledge the line between the latter and one-shotting is potentially diffuse and requires careful monitoring.
Any use of AI, and lack of AI use, should be carefully and precisely acknowledged in all papers going forward. In particular, the genesis of idea generation should be explained in detail. Exposition, both in written form and in presentations, should take a more important role than it currently has. Researchers should not be required to to use AI, or to use AI in a specific form, as a precondition for their participation in the mathematical community.
Given the high stakes involved and the rapidly evolving landscape, I believe mathematicians should take a more proactive role in helping shape the future of our profession. These are some current events I feel strongly about:
September 20, 2026: There is a new(ish) Association for Human Mathematics. It welcomes mathematicians who pledge either limited or no use of AI in their work. While it looks like my current AI use narrowly excludes me from joining, I think it is a very positive development as a variety of responses to AI developments will make the mathematics community more resilient and diverse.
September 11, 2026: 25 Fields medallists have signed a declaration with the title "A severe misalignment between AI and mathematics". I encourage everyone involved with the mathematics community to read the declaration and to consider endorsing it. (I don't completely agree with the contents of the letter, but I think it is important to make our collective voices heard and this involves making strategic compromises.)
There is a new Mathematical AI safety institute directed by Fields medallist Jacob Tsimerman and with involvement from other prominent mathematicians and scientists, including another Fields medallist and the president of the AMS. A priori, this could be a good initiative as AI safety could not be a more important and pressing issue, and mathematicians have a lot to contribute. However, I am very concerned by the way this initiative was launched. The center is planning on hiring up to 100 mathematicians in the Bay Area for at least a semester over the next year, which requires millions in funding. It presents itself as "Independent". Yet there is not a single word about its funding sources. Update (Sept 20, 2026): the MAISI website now discloses a $10 million grant from Coefficient Giving. While the silence at earlier stages is still concerning, I no longer have any objections
Below are links to essays, social media accounts, blogs and videos that I've found illuminating. They express a diversity of points of view and are from a wide cross-section of mathematicians, from students to fields medallists. (I can't keep up with the pace of essays on the topic so this list is becoming outdated.)
Proofs and prompts, a communal blog on Math and AI.
The big reset in mathematics by Daniel Spilman. This extensive essay is among the ones that come closest to my own current views.
A severe misalignment of AI in mathematics, by 25 Fields medallists.
Statement from Tristan Buckmaster regarding work related to Navier-Stokes .
Look, Mom, I pressed a button! by Danny Calegary.
The fate of the Riemann hypothesis, by Richard Schwarz. Short story about a dystopian near future.
@tao (Terry Tao) on Mathstodon. Extensive discussion on how the field should evolve, and lots more.
@littmath (Daniel Litt) on x . Lots of fantastic insight and reactions.
@thomasfbloom (Thomas Bloom) on x . Lots of interesting commentary, especially around Erdös problems (Thomas created the "official" list of Erdös problems and keeps track of which ones have been solved by humans and/or AI).
The last generation of mathematicians, Curt Jaimungal's podcast interviewing Jacob Tsimerman.
Tim Gowers's blog. Some very insightful posts on various topics around AI and math.
\begin{proof} . A declaration by a group of mathematicians who pledge not to use AI in any form.
The crisis of AI generated mathematics, by Max Weinreich. Another cogent case for refusing to use AI as a community.
Mathematics in the age of AI. Terry Tao's public lecture at ICM 2026.
No country for mediocre mathematicians, by Garvy . A substack post discussing (among other things) the contributions of "average" mathematicians as it relates to AI and more generally.