I knew that crowd during the era of Autodesk ownership.

It was a disaster. There were fundamental problems:

- Nelson had very specific ideas for a database architecture, and they were awful. Links going every which way, all of which had to be consistent at all times. Imagine URLs all having backlinks to all uses of the link. That's Xanadu. You could make this work today, but database architecture of the era was far too primitive. With all that linking, just rendering a page would take a lot of joins.

- Everything is pay per view in Xanadu. What are you going to store? Early thinking was to store the expensive newsletters which investors buy for a few hundred dollars a month. The people involved were way too into microtransactions.

- It was inherently a centralized system, like Lexis/Nexis, or the National Library of Medicine. This would not have scaled, because of the need to keep the backlinks consistent.

- No images. Images didn't fit Nelson's model of text editing. How do you charge for operations on images?

Probably the closest living descendants to Xanadu are the collaborative editing systems where multiple people can work on the same thing at the same time. The text systems for that are well known. There's even a graphical shared editing system, NVidia Omniverse Connector. Nobody uses it. [1]

[1] https://blenderartists.org/t/so-is-anyone-actually-using-omn...