A STATEMENT ON USAGE OF AI IN MY WORK.
This statement has been written, by myself without any AI usage, possibly awkward English included, and uploaded on November 16th, 2026. If it is still up and advertised on my website, and I am still alive when you are reading this, then I still uphold the views written therein, and any statements regarding AI usage in my mathematical papers is valid for papers written after that date. It is possible that my opinion will change as time advances.
IN GENERAL.
I believe LLMs have turned out to be great and useful tools for maths research, and I am enthusiastic at the fruits of my collaboration with them thus far. They seem profoundly helpful in bringing some ideas to life, some which I have already spent years thinking about; in speeding up some technical work, and tearing down barriers (such as: two communities using vastly different languages but using the same objects, making understanding the other community's articles difficult and unrewarding, for example).
It may be the case that AI becomes better at coming up with original ideas in the future, but this currently does not seem to be the case (caveat: as far as I can tell, with my limited skills in using AI tools, and the models that I have access to). I cannot claim that I will keep this collaboration-only mindset forever, as it could be the case that AI becomes able to write billions of excellent papers per second some time in the future, yet again uprooting any practices we may have gotten used to. As the evolution of that technology is essentially impossible to accurately predict, it is also an impossible task to prepare for the future; thus, I prefer to adapt and react to the current circumstances.
Our lived experience of mathematics is, ultimately, what interests me in mathematics: "living within" the objects, what images and emotions concepts evoke to me, the "technique" or "yoga" of working by hand with certain ideas... I do not believe AI usage, or the AI's existence, to be intrinsically incompatible with this experience of mathematics. Prompting and exploring ideas I genuinely like and care about, advancing my knowledge of them, is still "living mathematics".
I refuse to make any philosophical statements about the nature of mathematics with the authority of a practicing mathematician: this is far from my field of expertise. All comments above pertain to my lived experience of the job of researching mathematics. As a simple person, I quite like the view, beautifully explained by Lafforgue in this text (backup), that all things intrinsically deserve to be studied - even the most abstract and seemingly far removed from reality; and that all things hold infinite depth and beauty, and are infinitely intelligible. LLMs help us dig deeper into this well of intelligibility: they change the pace at which we do, which depending on one's opinions can be good or bad; they do not, again, threaten this view.
One aspect of our job is to be (and train) human beings able to understand high-level mathematical proofs. This by definition cannot be changed by AI. The intricate prestige system inherent to academic mathematics, which underlies all of the more prideful aspects of our job: competition to be the smartest person, proving well-known conjectures, attaining fame and glory, will likely die. How freeing that is! Bon débarras !
IN MATHEMATICAL ARTICLES.
I believe that retaining a human component in my mathematical papers is essential. For me, this will be upheld in practice by guaranteeing that I write all of my articles "by hand", myself, without any AI assistance, correction, rephrasing tools, etc. This means the "literary content", while the mathematical content will have been obtained with AI usage, the extent to which will be (for the moment) made precise in each individual paper. I think this "literary" content of our papers is profoundly human. I insist that this view is independent of the relative human skills to AI skills. While it is currently the case, I do not believe humans will remain better technical/matheamtical writers than AI's for long. As of writing this statement, AI-written papers are jarring to read, but this will likely soon change. I am also not claiming to be very skilled in writing. I simply prefer reading a humanly written paper with AI(-assisted) mathematics, than an AI-written paper with human mathematics: the worst of both worlds!