"Tech" and "Technology" on the other hand have a lot of overlap by definition, and a little overlap with, say, "coding" and "gadgets" and "3d printing." With user following, you have affinity. With tag following, you have taxonomy. Affinity is the subject of poetry. Taxonomy is the subject of doctoral theses. How are you doing the recommendations? Just thinking aloud, but it seems to me that if whatever analysis you're doing can "fuzz" tag boundaries the way it "fuzzes" your recommendations, if I subscribe to "#tech" and the commenters, sources and vote profile of "#technology" is similar, I should automagically get a few, some or all "#technology" posts as well. The onus should be on the system to provide the user with content, not on the user to determine the appropriate tag for submitting or subscribing. Take it one further - allow multiple tags, but give each succeeding tag half the weight of the previous. If I tag a post about Rupert Murdoch's twitter feed with #business, #media and #murdoched, anybody who subscribes to any of those three tags should get it in their feed. However, since it was tagged #media second, it should be half as likely to end up in someone's #technology feed than if it were ranked #media first. This gives the user an incentive to be thoughtful about his tags, rather than scattershot - by the time you get to the fourth tag you have 12% the influence of the first tag and by the fifth tag you're in the noise. Until, at least, you have several tens of thousands of users and an equal number of tags, and then that statistical noise becomes more relevant. Eventually, a solid subset of "tags" will build... but if people start tagging content with a new phrase ("#occupywallstreet") it begins "trending" just like a twitter hashtag. I'm no coder, but here from the cheap seats it seems like an approach that doesn't require tending, has room for growth and works the way I think you want it to work. Now go swimming.