Fireship · 2026-11-09 · notable
Fireship — 'OpenAI spent $1.2 million hacking Linux... and it was a bargain'
In short
Fireship breaks down OpenAI's Linux kernel audit, in which an unreleased OpenAI model came back with 948 root exploits for about $1.2 million in API credits.
Fireship · 2026-11-09 · 2 sources checked
A Fireship breakdown of OpenAI's agent pipeline that found 948 Linux kernel root exploits.
What is it?
'OpenAI spent $1.2 million hacking Linux... and it was a bargain' went up on Fireship on 9 November 2026. The video covers OpenAI's report 'A thousand fewer paths to root', where a model the public cannot use yet tore through the Linux kernel and returned 948 root exploits.
How does it work?
OpenAI's report describes a chain of agents: scouts picked audit targets, analysts searched them, triage agents reproduced bugs in a VM, exploit agents built working privilege escalations, and a final stage wrote patches for maintainers. The whole run cost $1,213,430 in API credits.
Why does it matter?
The title's 'bargain' points at the report's own maths: historical bug bounty prices would value the exploits at many times the run's cost. That gap is why defenders are being told to rescan old code as models improve.
Who is it for?
developers and security engineers