Files coming in would be immediately grouped by filetypes. My pictures/documents/music on steroids. Every different view would have its own relevant hierarchies and view styles. For most types of files, filenames are redundant, and they are better defined by their content and metadata.

Media files have names, producers and artists defined by metadata, there's no need for a filename. Documents have authors, titles and other info that can be inferred from the contents. Source code have class names and other kinds of  metadada like javadoc comments that can be used.

The music grouping could instantly group music by artist/album/song like in a music database, without that being some weird external effort. Any editing or recategorizing would be instantly reflected in the file itself, so the files stay in sync with what you actually see.

You should be able too multi-way tag and categorize files, no matter what their contents or if they support metadata or not. You should be able to tag your holiday photos not only by date, what holiday, location, but also any other sort of personal tagging you want. This can only be done by either an external database, or by a new "filesystem".

Any file entering the operating system should be immediately put into the proper location, music files should be placed in the massive music "folder", when downloading source code it should be added to the "source code locker", etc. If the system doesn't know what type a file is, it should look online to find available adapters, or ask the user to select what kind of file it is. In general, things should "just work", and you should never feel the need to designate "staging areas" where files just get dumped for "further processing", which most of the time never gets done. You won't need to end up with a "Downloads" folder full of files you might have needed at some point, but which you don't need anymore, but are too lazy to delete.