A few years ago I would have 100% agreed with you, but now with LLMs they may have made sense again.

yes, I caveat that at the end:

> Once we have strong AI, maybe the situation changes. But that’s what we’ve got today.

However, I am doing research on AI and hypermedia and I'm skeptical that the late-bound, discoverable dynamic of it compete well with things like MCP, at least if you are using something like a ReAcT loop.

I am publishing a paper on this soon.