I have been waiting a long time for someone to use lessons learned from recommender systems to build a social media site like reddit, but rather than having users group themselves up, they would be shown posts algorithmically, based on how they have voted in the past. For example, if I vote for a lot of pro EFF posts, maybe I will see more posts about free software or something like that.

This also allows the site to treat votes as merely an expression of preference, not a judgment of quality (which is hard to define). Many sites try to enforce a rule that says "Don't downvote things disagree with." This is an impossible battle to win, there will always be people who downvote things they don't agree with. This behavior needs to be accounted for in the platform.

This is more "socially scalable" to me.