I don’t quite understand all the downvotes unless people aren’t reading the post? I get that nobody wants a LLM to be writing drivers, but this article is basically a “let’s see if it can do it” - sharing the author’s experience and results (which he pretty much says not to use at the end)! It’s still an interesting read to see how the technology performs at this task and the number of hoops the author had to jump through to get any kind of functional result.

Letting AI code a kernel module is a terrible idea. A bad kernel module can easily crash the system and can create all kinds of security vulnerabilities.

All this does is make me not want to use FreeBSD.

No