Earlier quoted context omitted.
ChatGPT, Gemini, and Claude all point me to a plethora of sources for most of my questions that I can click through and read. They're also pretty good at both basic and weird git issues for me. Not perfect, but pretty good. Also part of the workflow of using AI is accepting that your initial prompts might not get the right answer. It's important to scan the answer like you did and use intuition to know 'this isn't ri…
_I_ know that the answer provided by the prompt isn't quite right because I have enough experience with the Git CLI to know that it wasn't quite right. Someone who doesn't use the Git CLI at all and is relying on an LLM to do it will not know that. There's also no reason for them to search beyond the LLM or use the LLM to go deeper because the answer is "good enough." That's the point of what I'm trying to make. You…
As is whether most documentation will be LLM generated in the future and what kind of quality that might be in 5-10 years. Right now, formal specs and documentation looks good for the first couple paragraphs and then is garbage.