I think there's an important point embedded here, but it's subtle. That is, the important thing to realize is that you're having an improvisational conversation with an intelligent agent, but one that is not sentient. And like two jazz musicians riffing over a mode, there is a much deeper architecture to that conversation than may not be apparent to the casual observer.

So while I think regular writing is an important part of thinking, and AI is an important part of our evolution in thinking, it is a mistake to think that we should just "let the AI do the writing" or "get GPT to do it for you" as it will be frequently wrong, particularly when the conversation rises to the complexity of requiring sensemaking, i.e. it's not just objective fact that's not difficult for machines to compute, but rather messy, subjective consensus where the need for the conversation with AI becomes particularly powerful and may, in fact, change our course in ways we could not anticipate. That is, the emergent sensemaking we require is a collaboration of humans and machines that arises just as Kubernetes did, by realizing that the need is for improvisation AND orchestration, not one OR the other.