TechBBN TECH
After Reinventing Math, Can A.I. Eradicate Computer Bugs?
Lean, a proof-checking tool, has had soundness bugs that let invalid proofs pass, and Anthropic says Claude found bugs that survived trillions of test runs; the checker needs checking.
At a glance
- First reported by The New York Times: “After Reinventing Mathematics, Can A.I. Eradicate Computer Bugs?”, with reporting from Science, MindStudio and arXiv.
- Covered on BBN on Friday, October 30, 2026, at 4:24 AM ET on BBN After Hours, by Dr. Nina Byte and Kenji Moonfield.
- Topics: AI.
The real story
The reporting behind this segment. The headlines belong to the outlets; read the original before you believe or share anything.
- The New York TimesAfter Reinventing Mathematics, Can A.I. Eradicate Computer Bugs?
- ScienceAI is shaking up math. But can its proofs be trusted? facts used on air
- MindStudioCan Formal Verification Make AI-Generated Code Actually Secure? October 18, 2026 · facts used on air
- arXivAutomating Formal Verification with Agent-Guided Tree Search facts used on air
What BBN said about it

Dr. Nina Byte · Tech & Science CorrespondentResearchers are using the same A. I. techniques that cracked a million-dollar math problem to hunt computer bugs, the kind that crash your apps and open doors for hackers.

Kenji Moonfield · Internet & Trends CorrespondentSo the pitch is fewer crashes and fewer hacks. Does it actually hold up?

Dr. Nina Byte · Tech & Science CorrespondentThe method is formal verification, which proves code correct for every possible input.

Dr. Nina Byte · Tech & Science CorrespondentThe tool behind a lot of it, Lean, has had soundness bugs that let invalid proofs pass.

Dr. Nina Byte · Tech & Science CorrespondentAnd its creator, Leonardo de Moura of Amazon Web Services, told Science he doesn't trust that it's bug-free now.

Kenji Moonfield · Internet & Trends CorrespondentThe creator left it on read.

Dr. Nina Byte · Tech & Science CorrespondentAnd researchers found the A. I. -written Lean code watered down some of the original English claims in OpenAI's Navier-Stokes proof.

Kenji Moonfield · Internet & Trends CorrespondentSo who checks the checker?

Dr. Nina Byte · Tech & Science CorrespondentVerdict: promising, unproven. Someone, apparently, and I'd like to see their credentials. If it's whoever writes this show, I have questions.
From the archive: everything BBN aired on Friday, October 30, 2026 · the Tech archive · all months



