He's right in that Facebook was helpful by auto-creating certain "groups" of friends for you based on profile information. Where this completely breaks down is once you move past the trivial task of auto-populating categories for your location, school, and work. As I see it there are three modes of sharing: The first is where my post is quite innocent and generic, so I just want to declare it to the world. This goes…
Reading your three situations, I have another one which I've been thinking about: a post which is informative and in some way beneficial to the public, but which a significant portion of the people you know simply do not care about. For example, you may have an excellent piece of commentary about the history of Unix and the state of X, Y, Z in modern operating system design. Most of your non-technical friends, along…
1 - determining the type of content based on words, size, etc. based on the type of content the system would automatically filter it on the receiver's end. if the receiver has typically ignored or deleted 'history of unix' type stuff from me, the receiver/reading app will file it away, and note to the receiver that something's been received, but it's probably not important. Like 'spam/ham' filtering, but as another dimension. (yam? tasty for some, untasty for others?)
2 - using signals of other people in the ecosystem as more signals in the 'spam/ham/yam' algorithm. If I post something that all my geek friends read/share/comment about, those interactions should count towards an 'interest' score on that piece when it's presented (or not) to my mom, for example.