Semi-automated organization is difficult, not because we don't have the algorithms to classify things into categories, but because everyone has a different idea about 1) which things belong where and 2) how much of it should be automated.

This is why everybody else's organization scheme feels alien to us, and why our own organization scheme, no matter how much sense we think it makes, will inevitably feel alien to everybody else.

If you build a tool to automate part of the process, and then release it to the whole wild world, people will complain that it files things in all the wrong places. If someone else builds a similar tool, and you use it, you will complain that it files things in all the wrong places. Because each of us not only has a different ontology, but also has different ideas about how we should approach ontology in the first place.

You could make your tool highly flexible so that it can accommodate all sorts of different ontologies, but then everyone will complain that there are too many options. It really sucks.

Oh well, I guess that's why we've ended up with a gazillion incompatible tools, each of which is probably good enough for whoever built it, but none of which really suits anybody else's needs.