[zahlman]

The point of an LLM is that it can take your problem as input, along with answers to previously asked questions (perhaps implicit) in its training data, and attempt to synthesize a solution to a problem as output. Here, a "question" is something that can be found with a search engine - something that directly presents the/a crux of an issue, which is identified after a debugging session (for a problem in existing code; for projects still in the design phase, a how-to question emerges after coming up with user stories and breaking tasks down into their logical steps). The point is that the question can be relevant to many different people who have written different code which encounters different problems - all caused fundamentally by the same conceptual misunderstanding (or non-understanding).

Stack Overflow is explicitly not designed or intended to solve problems or do the decomposition of the problem for you, nor the synthesis of answers. Because the result would never be useful to anyone else. The entire point is to have something searchable, and to allow answer-writers to keep their explanations DRY.

This has spectacularly failed, because no matter how frustrated people get with traditional discussion forums (https://www.explainxkcd.com/wiki/index.php/979:_Wisdom_of_th... among many other typical complaints), they apparently are much more suited to human nature.

Heaven knows how Wikipedia managed to avoid devolving into "Quora but even worse because you can scribble over someone else's post".