The essay is a first-person lament from a mathematician who sees recent AI-assisted counterexamples and proofs as more than a productivity story. The claim is that mathematics has never been just about getting correct answers. For many practitioners it is a human, social, even spiritual practice of discovery. If novel proofs can be summoned from a model "like DoorDash," then the activity that gave the field its aura and many mathematicians their identity starts to look hollow. The piece is not mainly about whether the results are useful. It is about what happens when the fun, status, and meaning were bound up in being the one who discovers.
Most of the high-signal reaction accepted that this is a real grief response, then sharply separated "mathematics" from "the profession of mathematician." Several working mathematicians said theorem production was never the economic center of the job. Teaching is. Publishing new theorems is the credentialing ritual that confers status and access, not the thing society directly pays for. That framing changes the threat model. AI theorem proving does not automatically erase the discipline, but it may scramble academic hierarchy, prestige, and the pipeline by which mathematicians earn a place to teach and mentor. Others pushed the same distinction from another angle. They argued that enjoying math does not require being first, just as seeing the redwoods is still meaningful even if millions went before you. In that view, better tools only remove literature search, case checking, and other drudgery.
The strongest optimistic case was not "nothing changes" but "the work moves up a level." People pointed to
Lean,
Isabelle/HOL, and
formalization as examples where AI is already useful for grinding through proof details while humans still choose definitions, spot patterns, reformulate problems, and decide what is worth asking. Several commenters said this could open mathematics to outsiders who lack departmental affiliation, library access, or years of specialized training. But that optimism came with a real caveat. If AI captures the enjoyable, legible part of the craft while wages and institutional slots shrink, then telling people to "still do it for fun" misses the point. A lot of the unease is about losing the ability to make a living from the activity that formed your identity.
The deeper line running through the comments is that this is not really a math-only story. Programmers, artists, and other knowledge workers recognized the same pattern immediately. People who love a craft are not soothed by being told they can now supervise more output. They care about doing the work, not just directing it. So the discussion landed in a place that is less about whether AI can prove theorems and more about what counts as meaningful human participation once systems can do a large share of the technically impressive part. The optimistic camp sees a dawn of abundance and new questions. The pessimistic camp sees "model babysitting," weaker labor leverage, and a fast collapse of the social bargain that used to fund mastery.
kirwinhampshire.substack.com