> So I don't see Edgerank as a band-aid fix, but a fundamental requirement for a large-scale system like this. Because (a) on the demand side, recommendations a

> So I don't see Edgerank as a band-aid fix, but a fundamental requirement for a large-scale system like this. Because (a) on the demand side, recommendations are what people actually want - let computers do what they're good at and reduce the human tedium of sifting through inane FourSquare checkins etc I can't speak for everyone, but I believe the Google Reader model would work for most people. The interface is entirely based on the source, not the post content, the keywords, or an algorithmically derived importance. There is nothing technical about it: you notice who posts interesting content and who does not. There is no need for automatically ranking the importance of two posts from separate sources, people are very good at deciding who they find interesting. Edgerank or proposed twitter feeds of the future are based on the idea of algorithmically figuring out what is the most important news for you right now. But if you don't care about learning about the "news" two weeks later, then a source-based manual filtering would let the user do that far better than a ML algorithm can. And if it's out of date in two weeks, it wasn't worth reading at all.