GHC tells AI-assisted contributors: disclose your LLM use, but don't let it do the talking
AI Contributions to GHC
GHC now has a formal policy for LLM-assisted contributions, built on Simon Peyton Jones's draft from the ghc-devs discussion. It sets four rules and one strong preference: prioritize people and community, make merge requests easy to review, require an identifiable human author who can explain every line, and disclose direct LLM generation via an Assisted-By line. Writing the final version yourself is preferred but not mandatory. The policy targets reviewer time, not model quality.
A patch produced in twenty minutes can still cost a maintainer two hours to understand.
- eschew
Seems ironic that a blog post describing GHC's AI policy, from a contributor to GHC, which expresses agreement with GHC's AI policy, itself directly violates GHC's AI policy (points P1 and P4)!
- memjay
Most important one is human communication.
Use AI as much as you want, for research, writing code, reviewing code, etc.
But please, keep communicating on a human level. We are slowly loosing valuable communities when only bots are arguing with each other.
- cadamsdotcom
To the authors: please include a few words giving context on what a GHC is so people don't have to seek that info out before trying to understand what you're saying
Didn't slow me down too much but was still a hassle.