Something has occurred to me a couple times when I’ve thought about Melkjug.
Suppose Bob has their personal Melkjug running, and one of the feeds Bob wants to keep track of is John’s Melkjug instance. In fact, he gives a pretty high personal priority to the picks from John’s Melkjug, because John has excellent taste in internets.
Meanwhile, John also admires Bob’s Melkjug feed, and unbeknownst (or perhaps knownst) go Bob, he adds Bob’s Melkjug’s feed to the big pile of links that his own Melkjug sifts through, and gives Bob some priority.
What happens in this case? What do we want to happen? Depending on the way Melkjug handles this, it could spiral out of control: Bob and John’s Melkjugs might go haywire and converge on an identical set. Or maybe our algorithm rocks so hard that instead we get some nice properties out of this, like a way of having the importance of a link rise with its social popularity in a way that mirrors Google’s PageRank, but is more personally tailored to ones social network. But regardless, if Melkjug is supposed to really become a tool for “collaborative filtering,” it would be worthwhile to think about how different users’ Melkjugs should interact with each other.
Some of our discussions have gotten a little mathy lately, so I’ve opened up a new section of the melkjug wiki, 
Melkjug sifts through a big pile of articles that you may want to read and tries to pick out a small pile of articles that has a good mix of characteristics that are important to you.