News for hackers. Stuff that matters.
Re:The "confident idiot" problem: Why AI needs hard rules, n... by max51 on Monday December 08, 2025 @02:14PM (#46196372)
>LLMs don't do this

They did at the beginning. It used to be that if you wanted a full answer with an intro, bullet points, lists of pros/cons, etc., you had to explicitly ask for it in the prompt. The answers were also a lot more influenced by the tone of the prompt instead of being forced into answering with a specific format like it does right now.

There are no comments on this comment yet.