The "Alan Kay" references in the video is a good cue of what's on the author's mind -- I guess the goal is to create an environment for contents, not something like a general-purpose desktop.
Quite ambitious I'd say! But I would also suggest that this goal (if I guess it right) could be better explained than just claiming this to be a "desktop reimagined", considering that a commodity desktop has neither a trackpad or eye-tracking device.
The discussion here has already made it clear -- hashtags for file organization is anything but new. It can be even simulated with a filesystem pretty easily. Just post everything into a big content folder, and then create a folder for each tag, then `ln -s`.
It'd be more awesome if a miner extracts semantic tags and metadata from the files. Not a new idea either, there were google desktop, gnome-tracker etc. macOS spotlight seems to be the most popular descendant these days. Microsoft is making another push with Cortana/MS Graph/Windows Search or whatever they call it.
The relationship between the contents is also an interesting aspect. Hashtags just throw things into hash buckets but do not help to relate in a broader sense. Even wiki links/hyperlinks are more effective. Project Xanadu also has interesting ideas that a reference link addresses by content so that you can reference a portion of the source (without the source defining them, unlike html anchors).
But again, this was the vision of the pioneers, but did not take off smoothly. IMHO partly because this idea is too big and involves a lot of smaller (yet still challenging) pieces (natural language understanding, resource description, app interop, and also, including the UX problem that the op is trying to solve) and none of the previous trials have both the depth and breadth to cover enough aspect of a user's everyday routine with a desktop environment.
Nyawww!