OpenAI 1000 Linux bugs for 1.2M - I don't buy it
OpenAI claims they fixed 1000 bugs in Linux and want to charge 1,200,000 for it. I don't believe a damn thing about it, for multiple reasons.
If Linux just waits they will get the model like always anyway, competition forces them to do that anyway. So they will do it themselves for 5000 in the future.
You can't just commit 1000 fixes into such a sensitive system, Linus is too based for that. Who is supposed to review all those fixes then? So many catches.
Let's just see if they get their 1.2M. Absolutely don't do it.
But they are under pressure now because it is a hostile marketing stunt by ChatGPT, because if they don't accept it then it looks like they don't care about security, which almost nobody understands a damn thing about and so it always creates uncertainty.
If that deal goes through, I rate the chance almost zero that it actually gets merged. The people who review also get paid and that is not free either. Nothing but problems. This is absolutely not friendly.
Well, let's see. Let's see if I am right about this.
Comments
Correction on my earlier post
I got this wrong. The sources - OpenAI's own research post “A thousand fewer paths to root” and the reporting based on it - say something different from what I wrote.
OpenAI did not claim they fixed 1,000 bugs and want $1.2M for it. They ran an internal research pipeline with an unreleased cyber model that found and developed working exploits for ~948 local privilege escalation vulnerabilities in the Linux kernel, plus candidate patches. The $1.2 million figure is their own API cost for that run, not a price they are charging anyone.
They are feeding the findings and patches to kernel maintainers through their Patch the Planet program (with human review on their side). There is no $1.2M “deal” for the Linux project to accept or reject.
I got baited by a scammy YouTuber who misrepresented the claims. Sorry for the incorrect take.