Some of our chats with AI can become training data. Even without another breakthrough in AI methods, a simple magic trick is possible: a machine connects ideas from mathematicians who have never spoken to one another and produces a solution. The humans were merely close; the machine solved it.
This is the trick of the psychic. This is the trick of Tom Riddle’s diary.
Martin Gardner’s Fads and Fallacies is a great book and worth revisiting. It summarizes interesting instances of pseudoscience and fraud. A cold reader can easily sell you a car. A competent hot reader, who benefits from information acquired beforehand, can convince you of just about anything with enough time. An audience, and often the psychic themself, that cannot see those sources can mistake access for extraordinary insight. A correct proof remains an achievement, but its provenance matters. We being read by AI, and subsequently our own utility is read for filth.
The funny thing about AI mathematics is that fear of being left behind encourages us to supply more of our work. Where those interactions enter training, fear of being scooped helps nourish future competition. The more you panic the more you feed the psychic beast.
Such coordination arbitrage has real utility. It also exposes our own failures. If we each learned an epsilon more about one another’s work, how many results would already have been proved?
Before declaring the mathematical community obsolete, we should consider what resources it has been givenand where future support is being directed. NSF’s new strategic plan for 2026–2030 explicitly emphasizes industry partnerships, startups, and market-ready solutions. What might comparable investment in grants, students, and sustained human collaboration accomplish over decades? We may be losing a generation of mathematicians to the fear that they will have nothing to do. My general belief is that whatever form mathematics is in, we will operate in that environment.
AI’s benefits can also percolate through human conversations. Universal adoption need not be the price of admission. For maintaining cognitive sovereignty, I would bet on people practicing moderation, skepticism, and having fun over performers dressed in off-the-rack theorems.
There are tedious tasks worth automating, and tedious systems worth abolishing. Perhaps we could write shorter papers and help one another expand them. There is plenty to be excited about. Before telling the diary everything, we might try talking to one another.
Many things seem inevitable, but we have a lot of agency over how things play out. If mathematics continues to have value, it will be because the community maintains its values. Mathematics has been a huge force for good in my life, and we deserve to share that with the next generation.
Received 22 September 2026.
Add to the discussion