There should be a permaban from HN for using LLMs or other AI to write comments, as they provide no useful, authentic discourse.
We should use AI to preload all of the boilerplate conversation for a thread first, then leave the few interesting comments remaining to human beings.
What if I had a HN plugin taught on my usual ramblings. If my post I was about to send matches what the AI already predicted what I was going say, it could flag me and ask "do you really want to do this?"
(Edit to add: this is not an invitation. Hope that works.)