No, it makes total sense: As I read it, this new AI policy introduces restrictions and additional burden on contributors, effectively excluding non-programmers that want to contribute AI-assisted code changes. The burden of proof is on the proponents of such restrictions to show that there is a neccessity for introducing such a change. It’s not the other way around as there is currently no AI policy with such restrictions (e.g. everything that is not specifically forbidden is allowed).
I really don’t agree with this. LLVM is, at its core, a compiler framework that’s responsible for the running code on the majority of the world’s computing systems. From embedded devices to servers and everything in between. Like the Linux kernel, quality and stability are IMO the highest priorities here. We don’t need more contributors who aren’t programmers to contribute code. That may suck for those people who want to feel good by having some code accepted into a prestigious project like LLVM for their resumé without learning compilers, but I couldn’t care less about those people. Those people can take their contributions to one of the other million OSS projects, LLVM will be just fine without them.
What we need more of are contributors who’ve built experience, can review code, learn and teach others through discussions. We need people with good taste that comes with years of interaction with LLVM. We need people who can look at a PR that seems to do a correct thing on a local level, but explain that it’s not the right place to be making those changes in the long term.
To sum it all up, I strongly agree with this RFC and we can evolve it as needed in future.