OpenAI's latest data dump shows AI solving math problems, fueling job fears

The leak of a set of 377 math problems solved with AI has reignited the debate about replacing skilled work. There is no evidence that the mathematics profession is dead, but there is evidence that certain tasks are being automated fast.

The leak of a batch of **377 math problems** solved with artificial intelligence has reignited an old debate: what happens when a machine does, in seconds, what takes a professional hours? The short answer is that it depends on which part of the job we're talking about.

## What has and hasn't been leaked

The material circulating consists of **377 math problems** solved with AI. It is not an official exam or a peer-reviewed evaluation: it is a set that has been made public and is used as an informal test of capability.

Jumping from there to the conclusion that mathematics as a profession is disappearing is a huge leap. Solving closed problems and verifying proofs are different tasks, and the second still requires human judgment.

## Automation doesn't erase professions, it erases tasks

Economic history is rough on catastrophic predictions. ATMs didn't eliminate bank tellers; they relocated them. The spreadsheet didn't kill accountants; it took away their mechanical work.

Something similar is happening with mathematics. AI speeds up calculation, pattern search and partial verification. That can reduce demand for certain junior profiles and increase it for those who know how to frame problems and audit results.

## The antiestéticar changes sides, but the data doesn't back it up

Some claim AI will leave entire nations sweeping floors. That's a hypothesis, not a fact. Employment data in technical sectors do not, for now, show mass destruction linked to these tools.

What they do show is pressure on repetitive tasks and a rising demand for judgment. Whether that translates into layoffs or retraining depends on business decisions, not on the technology itself.

## What to look at before declaring a profession dead

Three things: whether the task is closed or open, whether the result can be verified without a human, and whether the cost of error is high. In applied mathematics, all three weigh heavily.

Until that is clarified, it's worth remembering that no set of **377 problems** proves on its own the end of a profession. Caution, here, is the only thing that can't be automated.