I find it so bizarre, too. I’ve been using quite a similar autosuggestion feature as part of Fish shell for a few years now. But when an LLM keeps spewing words at me, that’s a whole different shtick. It genuinely just inhibits my thinking, which is a feeling I never had with Fish.

I guess, one difference is that Fish uses real intelligence, a.k.a. my shell history. If it has a suggestion, the chance is high that it’s actually what I want to do or close to it. And it also shuts the hell up when there’s no good suggestion. I don’t have to be constantly vigilant that what it suggests might be complete garbage.

And the other difference is probably that it’s *my* intelligence, *my* shell history. I will have thunk the thoughts before which lead to the command it suggests, which brings the brain load much further down again.
Occasionally, it’ll suggest something where I have no recollection of having run that command before, but knowing that I have, is still really useful and this only happens for niche commands anyways. Most of the suggestions are just stuff which I’ve run a few minutes ago or last week or such, where I won’t have to think about it.

I guess, it probably also helps that commands have simple formatting, with only a single line and you can mostly read the flags in any order…