Solution: All good journals have RSS feeds and in the context of academic interests you can "easily" detect relevant papers based on author, keywords and affiliation (one might eventually add some social feature as well, but it definitely would work based on individual ratings without any social features). So a user would select some journals and starts by staring the relevant papers or authors or ... After a while there would be enough evidence accumulated to support the classification of new papers into relevant or irrelevant groups.
Also there would be room for all sorts of alerts and sharing and commenting and ...
I (and I am sure many others) have had this idea for a long time but I am a currently not a position to do this myself and other people who attempted to do it had done it wrong, I was wondering if there is anyone who want to do this (I would be happy to help in design or even implementation or I'd be also happy if anyone gets the idea and build their own tool without getting me involved).
Ali