The basic reason why we don’t have the systems envisioned by Kay, Nelson, Engelbart, etc. is because we have ignored or forgotten the components of the systems those folks actually built that make what they envisioned possible.
Can’t have bidirectional visible connections without permanent addressing.
Can’t have dynabook-style experiential learning without open-ended interactivity.
Can’t have symbiosis between a computer system & a professional community without the kind of mastery of tooling that downward-sloped learning curves produce.
Can’t have properly useful user agents without declarative programming & planners.
It’s not like these things are impossible to implement. They are 70s tech, or 60s tech, or just a particular mental frame. That’s why these systems could be created in the 60s and 70s.
Because they could be made to work in the 70s, they could be made blazingly fast today (with fewer developers involved, cheaper hardware, & no institutional support).