"twitter is largely a one way medium. Celebrities -> everyone else"

"I'm not convinced that this is true. Here look at this random kid I picked. Nearly everyone he's following looks like a peer:"

This is the problem with inferring statistical properties from a sample size of 0 or 1. Here's some data on the subject: http://www.intmath.com/blog/twitter-follower-semi-log-graphs...

The sharp drop from having 0 to ~1000 followers on the graph on number of followers indicates that twitter is somewhat of a one-way medium.

I don't think that data supports the claim that Twitter is a one-way medium. What it shows is that there are a significant number of Twitter relationships which are one-way, but that doesn't mean that any individual is solely using Twitter as a receiving medium. What you would need is data on the number of users with no followers who are still reading their timeline; in the linked post, Dalton says this is high, but doesn't provide a link and I can't find data on this.

I actually think both of you are right. It's not about exclusively communicating with your friends or exclusively following celebs, it's about both. What you care about combined with a personal audience to discuss it with. The kid you found as an example subscribes to Paul Ryan's and Newt Gingrich's feeds along with his friends' feeds. Twitter can target Republican, right-leaning ads towards him. Maybe he discusses politics with his friends online and off.