I fully agree with you Julian, and until we have proof that there are large-scale cases of calculator-induced psychoses and cognitive degeneration I'd like to suggest that we stop with the false equivalences. LLMs are a powerful tool which means that, much like a power tool, it must be handled with care and safety precautions. Otherwise you're just going to harm yourself and possibly others. On a personal touch, I worry that contributors who absolutely rely on LLMs will be excluded from participating when the token price ceases to be subsidized, and AI companies are brought back to reality and must make profits, if not break even. Another reason why cultivating the contributors is more important that getting code merged at any cost. Le 16/07/2026 à 14:25, Julian Ospald via ghc-devs a écrit :
No one in *this* entire thread has been suggesting to neglect a patch purely on the basis of LLM use.
The policy very clearly argues against this.
Although often mischaracterized as a loud minority, there's a fair number of people who were in principle pro a blanket LLM ban. You're talking to them right now and if the policy is watered down to "well, quality" I don't see what any of us gets out of this. Why should we meet you half-way?
Other communities (see Agda) have effectively split over this. And this thread really makes it look to me like any concerns on LLM use or additional burden on LLM users is seen as an attempt to "segregate", shame or violate those peoples privacy.
I find this line of argumentation to be false and actively misleading.
Even the current AI policy demands additional burden on LLM users. It is absolutely reasonable and common sense, unless we go back to comparing them with calculators and ignore the large body of scientific evidence of risks and the massive negative fallout in other open source projects.
In order to salvage *some* positive use, we have to be mindful about its use, not ignorant. This is not discrimination, this is caution.
On Thu, 2026-07-16 at 17:17 +0530, Harendra Kumar via ghc-devs wrote:
I am sure there are many others like you and they are very capable potential contributors. We do have opposing views in the community but I do not think anyone should feel discouraged to contribute because of those who hold opposing views, it will be detrimental to the project, I am sure they will find enough supporters. The real problem starts only if an otherwise good patch is rejected or totally neglected purely on the basis of LLM use.
-harendra
On Thu, 16 Jul 2026 at 14:58, Tom Ellis via ghc-devs <ghc-devs@haskell.org> wrote:
On Thu, Jul 16, 2026 at 08:44:10AM -0000, Julian Ospald via ghc- devs wrote:
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. To add my personal point of view on this, it is only the advent of LLMs that has made me think I could possibly become a contributor to GHC. 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. (The same goes for most projects actually.)
Tom _______________________________________________ ghc-devs mailing list -- ghc-devs@haskell.org To unsubscribe send an email to ghc-devs-leave@haskell.org
ghc-devs mailing list -- ghc-devs@haskell.org To unsubscribe send an email to ghc-devs-leave@haskell.org
ghc-devs mailing list -- ghc-devs@haskell.org To unsubscribe send an email to ghc-devs-leave@haskell.org
-- Hécate ✨ WWW: https://glitchbra.in RUN: BSD