Yep. This is what I'm trying to beta test with stuff like #steemit-hunt. I think we can (for now) just make do with having groups of people be thoughtful about using tags and self-policing.

Then we can also build alternate client front ends which are optimized for different kinds of use cases.

Of course, Hive account@jesta's #steempress project might very well be used to have individuals RUN those separate "sub reddits" in even more of a distributed way, even with separate domains.

It does bring up the issue of trust. What to display / what not to display / algorithms for content sorting can very much influence what people see, which influences how they vote.