[way66]

Foursquare has a API method that exposes the people who are at the same place than you are. Then we get their twitter nicknames and find who you have the most followings in common.

We use Rails 3, Memcache, Redis, Resque, Foursquare API (with Push) and Twitter API.

[sfphotoarts]

Thanks, that's essentially what I said, but I am curious what you mean by "most followings in common" (you mean people intersect?) I assumed you did some form of semantic analysis of the tweet streams to intersect based on common RT's or link topics?

I'm actually surprised there is a maxima between people intersects, or that it carries some meaning.

I wasn't really asking about the front-end, that seems like any of the boilerplate environments would have been equivalent, and just becomes a matter of most familiar, but rather the 'backend'.

Let's say you can write a UI, you have a lat/lon, you have the tweet streams from those around you, how, in a meaningful way do you find similarity? That's the core of your idea. Because Mike and Brett both follow JetBlue what does that tell you? However, if Mike and Brett both RT'd a link about the speed of limit hard limit, it's reasonable to assume they have some similarity in interest. That's the part I was interested in getting your thoughts (mainly because its something I've both thought about and worked on)

[way66]

Try to run some computation on top of Twitter using the followings in common and you'll see that it's actually working great. We only display with whom you have more than 4 followings in common.