I'm thinking that it may be intersting to have a filesystem that is actually a database system. The fs would store metadata about the object and then be able to search for the file based on that metadata. Also there is a possibility for the fs to make smart views. So automatically make a directory for the band radiohead when you dump in a bunch of radiohead mp3s. There could potentially be a way of searching via the path and getting a directory that would have the results of said query.
My inspiration here of course was the mp3 thing, but of course this could expand more generally. You may have documents say meeting minutes that you want to keep track of. Only the past month are really intersting to you so you create symlink from /work/meetingminutes/search/age<=30days/ to /work/RecentMinutes/
I'm going to toy around with this to see what comes out of it.