When mathematicians were human

The book When Computers Were Human (Grier 2005) describes how a skilled craft was invented, evolved and grew and then was swept away by the first few generations of electronic computers. When the tide receded, what remained were the even more skilled mathematicians who programmed the computers.

Are those mathematicians about to go the same way? In September, OpenAI announced a partial proof of the Navier-Stokes problem, an open question in the ‘Fermat’s Last Theorem’ class of difficulty and importance. That caused some concern in the mathematics community, not least because of suspicions that OpenAI was free-riding on work nearing completion by humans.

That has been eclipsed by OpenAI’s release, on 6th October, of 722 papers claiming to settle 372 open problems. The release can be found at:

https://github.com/openai/math/

Some immediate responses from mathematicians surveying the results are at:

https://proofsandprompts.com/2026/10/08/100-reactions-to-100-solutions/

While it is too soon, just five days after the release, to say that the craft of pure mathematics done by humans has been superseded, it is clear that something startling has happened. Quite a few of the responses mention feelings of shock, grief and mourning. No doubt Adam Smith’s pin-makers felt grief and mourning too, when their craft of making lovely pins was snatched away by manufacturing. Hard-hearted economists (and funding bodies) might not have much time for that. Even they might feel sympathy, though, for the younger people who have devoted the best decade of their lives to reaching the frontier of knowledge, only to see it accelerate away. Who would now contemplate spending years on a hard problem? Worse, if you happen to be the head of a maths department, who would now embark on a maths degree? What will the school teacher say to the reluctant pupil who asks “please Miss, why are we doing this?”

The dust has been raised, it has hardly begun to settle. The word ‘slop’ also appears quite often among the responses. Some of the papers are described as unreadable and it appears that some have already been withdrawn. A mountain of work faces the unfortunate humans who will have to decide on the quality of these outputs. For some it may be their last contribution to mathematics, for others a nervous reprieve. Though many of the papers may be flawed, if enough are correct, the craft of pure mathematics may have to change.

One may ask what OpenAI’s intentions were, other than sheer impact, and whether they had really thought through the consequences. Or was it mathematicians who thought, as the pin-makers of old maybe did, that some things would never change? They can perhaps console themselves with the thought that human computers have gone, but more humans than ever are involved in controlling what electronic computers do. For now.