I think that dismissing it out of hand as useless is unproductive.
From my understanding Roy Fieldings dissertation was a description and formalization of the web as it has been built, not just prescriptive advice on software architecture. To say it's not useful is to ignore how humans browse the web.
my memory is hazy but aren't semantic web, ontology and all that HATEOAS stuff for bots instead of humans? Now that LLM is here, what use do they have?
I haven't really thoughy about REST in relation to all the recent ML advancements. REST (With HATEOAS) might actually be a good way to expose APIs to these agents since they now have some ability to "reason" about responses beyond a hardcoded set of handlers
Structured data doesn't magically go away because you have better interpretation of unstructured data that doesn't blow goats; it's still orders of magnitude easier to manage, store, and retrieve structured data.
the web is not data, this is the fundamental thing that's problematic about that way of thinking. We already have structured data, it's called database
