On Mon, Jul 27, 2026 at 11:57:49AM +0200, Hécate via ghc-devs wrote:
In your case I'd say it's even more important that you do *not * use an LLM to give you answers about questions you may have regarding GHC's internal, as you will be unable to detect if they are correct until after it's too late (or someone else points it out).
I think you are raising a valid risk. In my particular case I am confident enough in my use of the tool that my use of it will help rather than hinder in that particular endeavour. I think your "you will be unable ... until after it's too late" is too strong a claim. I can only speak for myself rather than others. To others, I would suggest that, rather than avoiding using such a tool because of that risk, that they instead learn to use the tool in ways that will be an overall benefit rather than a detriment. If, despite their efforts, they decide that using a tool can only be overall detrimental to them, so they will not continue using one, then so be it. Tom