Disagree. I can be more or less certain that documentation authors or write-ups have knowledge about the topic im researching to a reasonable degree, because 1) they need accurate and efficient code that makes sense to establish a relationship with the reader and compete with other solutions 2) humans are bad at explaining things they dont know about opposed to executing or managing unknowns

ChatGPT? None of these. Its a giant probability calculator that wants my prompts and its own answers to foster itself. This also nurtures its inaccuracies and makes it easier to confidently lie. I hardly find any reasonable use for ChatGPT other than quick completion suggestions for at a maximum of 2-3 lines, because thats what its designed for.

You know that the GPT is trained on data including that documentation, right?

If you don't like ChatGPT, that's fine. But it's clearly useful to many people, and you aren't entitled to feel superior for not using it.

> you aren't entitled to feel superior for not using it

I will break the HN spirit but you kind of wanted to say that there. I didnt imply that at all, nothing near that. The dataset is just, imo, too big to give extremely accurate (or more accurate than human-thought) answers for specific questions.