> I wouldn't have the stamina to remain engaged otherwise. A > policy that discourages LLM use, in effect if not in word, makes it > less likely I will ever contribute to GHC.
Although I have contributed to GHC in the past, the above applies to me as well (pending an "again"). Contributing to GHC does not feel like an efficient use of my free time unless I instruct an LLM to carry out the designs I have in mind. The problem is the design that you put the LLM on; it is not the output it produces that needs continuous iteration by the (human) author. Similarly for MR descriptions: If you just take the output of the LLM unmodified you end up with a huge blob of prose that nobody wants to read. It is the same for an MR description produced by a human that has no clue about what they are talking. My eyes tend to glaze over. Same for an overlong mail thread. The problem is the output, not who produced it. Cheers, Sebastian Am Do., 16. Juli 2026 um 11:29 Uhr schrieb Moritz Angermann via ghc-devs < [email protected]>: > Magnus, > > I see you did not read my email, but got hung up on the word[1] I chose. > It is not about the mere disclosure. It is about the classification of > contributors into two categories, those that write code by hand (who are > strongly preferred) over those that use assistive technologies to write > code. > > Now, please explain to me how I am supposed to feel about this, as someone > who considers themselves in the second category? And what word choice you > would prefer me using? > > Best, > Moritz > > — > [1]: https://www.merriam-webster.com/dictionary/segregate > > On Thu, Jul 16, 2026 at 4:02 PM Magnus Viernickel via ghc-devs < > [email protected]> wrote: > >> I find the framing that this policy is used to "segregate" (given what >> the term usually refers to) developers to again be crossing a line. Nobody >> is threatening anyone's existence or dehumanizing them by asking them to >> disclose if they used an LLM to write their code. Since this is the second >> time this happened during this discussion, I would really hope we could be >> a bit less careless with our words. >> >> Best >> >> Magnus >> On 7/16/26 10:51, Moritz Angermann via ghc-devs wrote: >> >> Julian, >> >> I'm not sure how many more times I need to write this: It is >> fundamentally against my nature to tell anyone what to do, that includes >> telling people to use LLMs. >> It is probably from the same foundational values, that I am against >> telling people what NOT to do. You of all people probably know this even >> the best. >> >> You explicitly want a policy that puts judgement on people who use LLMs, >> and segregates the pure (hand written) from the impure (assisted) people, >> yet you >> claim you want to keep _all_ of them? You must see how qualified >> segregation tells one group that they are NOT welcome? >> >> > If we don't express a preference for human authorship, we're >> effectively saying "you have to use LLMs so stay relevant". I think this is >> very clearly your opinion. It is not mine. >> >> This is NOT my opinion, and I also do NOT think this logical conclusion >> follows. >> >> Best, >> Moritz >> >> On Thu, 16 Jul 2026 at 15:44, Julian Ospald via ghc-devs < >> [email protected]> wrote: >> >>> > Well, if you want GHC to become a political vehicle for an ideology >>> >>> Not spelling out values doesn't mean that you don't take sides. Your >>> constant "LLMs are here to stay" and "it's just a tool" is exactly doing >>> that, you are implying that: >>> >>> - you can't resist, stop resisting >>> - you have to follow everyone else or you will be left behind >>> - your value is only your output >>> >>> It is a mechnical view of open source and collaboration that is focused >>> on the product, not on the culture from which those products arise. My view >>> is different... I think it is the human culture from which those ideas and >>> products originate and Haskell has given a home to many engaged and curious >>> people. We want to keep those people, whether they use LLMs or not. >>> >>> If we don't express a preference for human authorship, we're effectively >>> saying "you have to use LLMs so stay relevant". I think this is very >>> clearly your opinion. It is not mine. >>> _______________________________________________ >>> ghc-devs mailing list -- [email protected] >>> To unsubscribe send an email to [email protected] >>> >> >> _______________________________________________ >> ghc-devs mailing list -- [email protected] >> To unsubscribe send an email to [email protected] >> >> _______________________________________________ >> ghc-devs mailing list -- [email protected] >> To unsubscribe send an email to [email protected] >> > _______________________________________________ > ghc-devs mailing list -- [email protected] > To unsubscribe send an email to [email protected] >
_______________________________________________ ghc-devs mailing list -- [email protected] To unsubscribe send an email to [email protected]
