Axiom Math wants to keep AI slop out of mathematics

Bulletin. Telescope log, Specola Galactica.

TechCrunch reports that Axiom Math, a startup, has deployed AI agents to formally verify the Stacks Project, described as one of mathematics’ largest reference libraries. The outlet frames the effort as a response to a broader problem it calls AI-generated “slop” in mathematics, unchecked or unreliable proofs produced by generative systems and circulated without rigorous confirmation.

According to the report, Axiom Math’s agents completed a formal verification pass over the Stacks Project’s contents, a step intended to confirm that the material’s proofs hold up to machine-checked scrutiny rather than resting on informal or unverified reasoning. The company is characterized as betting on verification as its core approach, positioning formal checking, rather than generation alone, as the safeguard against error entering the mathematical record.

The report does not detail the verification method, timeline, or scale of the Stacks Project beyond calling it one of the field’s biggest libraries. No figures on error rates, proof counts, or comparison benchmarks are given in the available material.

TechCrunch’s piece treats the milestone as evidence that AI systems can be directed toward confirming existing mathematical work, not merely producing new claims. The source material stops short of further technical specifics, and this bulletin reports no more than what TechCrunch has published.