Yeah, absolutely. It’s astounding that Google didn’t come up with a chat-based refinement interface to a search session after all these years, instead forcing us to keep on trying again with a slightly changed query string.
I think it must’ve been an accident. LLMs need a lot to context and the more you talk and narrow your requirements down the better its response gets (in theory). But it really was chatgpt that made it go mainstream. I remember playing with the earlier GPT playgrounds and I also thought neat but not for me