Also, the blogpost isn't the policy. The policy lives here: https://forge.rust-lang.org/policies/llm-usage.html
It's pretty explicit about what is and isn't allowed.
Also, the blogpost isn't the policy. The policy lives here: https://forge.rust-lang.org/policies/llm-usage.html
It's pretty explicit about what is and isn't allowed.
@slomo @zeenix Here's an example where you're allowed to use an LLM:
ℹ️ Changes are trivial if there is no other way to write them, or the other ways to write them are nearly identical. For example, the following are all trivial:
- Typo fixes
- Markdown links
- Changing a word to a synonym
- Type signatures for a trait implementation
So, LLM allowed, but only with disclosure
@diondokter @zeenix Yes, I read the actual policy too. That's equivalent to basically not allowed :)
It's pretending that there is some kind of compromise but that's actually not the case. Nobody* is going to use LLMs for such trivial changes and if they did the result would be indistinguishable from using other, more simple tooling (e.g. a spell checker, or sed) so there's not even any point in disclosing the use of an LLM.
That's perfectly fine FWIW. Their choice. But it would be nice to be clearer about that.
* Nobody might be a bit strong. I saw PRs of people fixing typos with Claude Fable. But that's probably more a sign of someone being confused than normal practice.
Also judging from my other question about another aspect of the whole thing, I think the policy is unclear to others too.
https://toot.cat/@slomo/117042581420327383
The answer to that is: this doesn't apply to security vulnerability reports and those are handled by the security team. Also unclear from the policy, and nobody could directly answer that either.
@slomo @zeenix Seems like that should be added to the scope maybe: https://forge.rust-lang.org/policies/llm-usage.html#scope
@slomo @zeenix It also leaves room for (private) LLM review. Maybe it spotted an off-by-one error and inserted the `+ 1` for you. You don't need to manually type that in.
But I agree it's largely a clamp-down on (mostly agent-based) LLM use.
But also, it's a big project without BDFL. So this is the policy that ultimately got agreed on. Fully banning wasn't accepted and fully liberal wasn't accepted. This policy is the compromise between both sides
> It also leaves room for (private) LLM review
That's the other part. It explicitly allows private LLM usage. The Rust project is not Apple, they don't have the means to tell people what they do privately on their own computers and definitely not the right :)
> This policy is the compromise between both sides
I don't see where the compromise is. My guess is rather that the pro-LLM side just gave up.
@slomo @diondokter yeah, I don't see a compromise here either. It's essentially an outright LLM ban. People don't need your permission to use LLM privately and there's no way in hell you can police that.