[fezzez]

Wow thanks for the kind words! I'm one of the makers. We still haven't found a solution for some of the problems that you've mentioned, but our focus from the start will be around privacy. I think you've grasped that problem yourself when you mentioned a self-hosted version.

On the technical side, we currently see this as something that can index on-the-fly, and maintain an internal representation of all the people who have the authorization needed to see the document.

I really like the idea of "auto-tagging". There's already research done on algorithms that generate tags from bodies of text. So it shouldn't be too difficult to integrate that into indexing.