I’ve been playing around in Persai‘s beta for a while now and wasn’t initially sure what to make of it. It’s a content recommendation site, pure and simple, and I have so much other content flying at me that I didn’t see how the site could fit into my daily grind. But the algorithms behind it prove more intriguing as the days go by.
There’s no learning curve with Persai: enter terms you’re interested in and it returns articles on a regular basis. No social networks, no friend lists, no ratings. Just news. So simple, it’s almost jarring at first glance. Once the engine returns relevant articles, you have two choices – read or reject. Both actions help the engine learn, yes, but Persai’s creators – Matt Kent, Kyle Shank and Ted Dziuba – don’t want users to set out with the aim of teaching the engine. “Just use the thing,” they say in the company blog. So that’s precisely what I did. Among the several interests I set up – including emerging tech, election 2008, independent film and semantics – Persai hits closest to the mark for election ’08 and semantics. I attribute the former to a large pool from which to draw and the latter to the specificity of the term. The independent film category has been the toughest for the engine to nail; for some reason, it keeps giving me articles about the Indian film industry. Not that there’s anything wrong with that.
I’ve never been in a beta quite like Persai before. Its creators don’t give interviews, there’s no whiz-bang design to pretty up the site, and I don’t think I’ve ever received an email from them, a la, “Here are our new features! Tell us how it’s going! Are you happy? Let us know if you’re not!” They’re not marketers, in other words. All that’s left to contend with is the technology and I find that profoundly interesting. Users are essentially thrown into the engine with little background or direction as to how to use it. I suspect this is what a beta looked like way back before it became a marketing term: wrestle with the tech for a bit and here’s a tiny Feedback link if you have suggestions.
Persai is in a profoundly tough space, considering the breadth and depth of companies playing in recommendation, machine learning, and news filtering. Its revenue model currently consists of contextual ads, which I’ve found to be quite close to the mark when filtered through their engine. In short, Persai has wormed its way into my daily grind, whether I wanted it there or not. I find articles that I’m not seeing anywhere else. The company is clearly smart about its algorithms. But how they push those out to the market, make them friendly for mass consumption and position themselves against weighter entities will be just as much, if not more, of a test.
