Good article. Not sure why Lobste.rs would have removed it.

One passage I wanted to comment on:

i don't know how more people don't experience this, but staring at an llms output be confidently wrong eventually to get there in the process, or looking at the commands its running and see some really stupid things its doing drives me crazy to the point i will just stop it before it answers, i cannot handle something being that confidently wrong (even though it'll probably still get there in the end)

Honestly, any time a piece of software causes problems/inconvenience/errors by "being helpful", I get irrationally angry about it, and always have. Autocorrect, autoindent, autocomplete, type ahead suggestions, anything like that. Any time something does that to me, I turn it off immediately. Red squigglies to indicate misspelling is fine. That's giving me information so my human brain can make a decision on it. But taking action for me and doing shit I didn't tell it to makes my defenestration reflex tingle mighty hard. Even if it does happen to do something that saves me a step, I still just loathe it immediately.

LLMs are the same thing, just distilled into its purest and most insufferably potent form. Words cannot express how glad I will be when this AI bubble fades into the kind of vague retrospective embarrassment we currently reserve for NFTs and Beanie Babies.