On Mon, Jul 27, 2026 at 07:53:04AM +0200, Sebastian Graf via ghc-devs wrote:
> Just want to chime in and say that Moritz is not the only contributor who
> wants to use LLMs in the future.

Holding up my hand as another example.

I'm not actually a GHC contributor but I would like to be in the
future. The only way I will be is if I am allowed to use agentic
coding in my contributions.  In fact the same goes for almost any
contribution I make to any codebase, open source or otherwise.  My
life is simply too short to not to use tools which reduce the effort
expended to achieve a particular result.  If I'm not allowed to work
in my preferred way in GHC I'll direct my efforts elsewhere.

By way of example, here is an ecosystem contribution (bug report with
reproducer and precise location of where the fix needs to be applied):

https://github.com/qrilka/xlsx/issues/205

The bug was discovered from a much higher level bug report entirely
but an analysis carred out by a coding agent.  The precise location of
the bug was discovered by the coding agent. The minimal reproducer was
written entirely by a coding agent.

I had to patch this locally as part of my job.  Whether I reported it
upstream was up to me.  My view is that the pro-social choice was to
report it upstream so it could be patched to everyone's benefit.  I am
glad the maintainers of that package does not have a policy that
discouraged me from making that contribution.

Tom
_______________________________________________
ghc-devs mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to