I just don't even understand why any kind of functionality like this is desirable to someone. Like, hasn't the dust settled now, hasn't the hype wained enough,

I just don't even understand why any kind of functionality like this is desirable to someone. Like, hasn't the dust settled now, hasn't the hype wained enough, and we all understand, for the most part, the broad and yet also weirdly specific utility of models like these? The whole plugin thing in general feels so dissonant in relation to the careful and couched copy we get from OpenAI about what these models are and are capable of. Like they want to say, for very good reason, that these models are a certain kind of tool with very real limits and huge considerations on safe, sensible usage. You can't necessarily trust it, it does not "know" things, and it is influenced by lots of subjective human tuning, blah blah. But then with all this plugin stuff they seem to be implicitly saying "no, actually you can trust this, in fact, its like a full-on AGI assistant for you. It can make PRs, directly orchestrate servers, make appointments for you, etc." Maybe I just don't understand?