Google Reader started offering a new sorting method for blog posts: Sort by magic.

While many tech evangelists have already declared RSS-reading dead and migrated to Twitter-style micro-sharing, I still think RSS is by-far the most effective learning tool/platform on earth (maybe on par with the courses I am taking in Stanford GSB).

There is a personal dimension for me in seeing Google Reader implementing post recommendations. About a year ago, I was working enthusiastically with two classmates to crack the personalized news problem. After putting 200+ hours into it and building a prototype, we realized that there were two most serious questions we needed to answer about our strategy before moving on:

1. How do we solve cold-start issue?
2. What happens if Google starts personal recommendations?

The first question was about people starting to use a new service (probably switching from Google Reader) and having high expectations from day 1. The cold-start issue was about the fact that we wouldn’t know anything about the user’s reading preferences until we collected his reading patterns. Therefore, during that "cold" period, the user wouldn’t get any better experience than what Google Reader offered. Plus, as he was "used to" a certain RSS-reading interface, he would inevitably have an inferior experience as all humanbeings have strong anchors to what they are used to. In fact, by spending so much time on cold-start issue, we realized that many start-ups were using social recommendations (I get recommendations from the posts Mike has read, because the engine thinks Mike has similar taste to me), because when you base the engine on social similarity, you can shrink the cold-start period close to null.

Second question was a preemptive critical thinking about what our competitive advantages would be, if any, in case Google starts serving personalized recommendations. First, they had the sizeable user base to "learn" how to make better recommendations. Second, they would already have vast amounts of reading patterns, by-passing the issue of cold-start. Third, the recommendation engine would have lots of synergies with their search-ranking algorithms. Among many additional dimensions, they clearly had the right to win.

Based on these concerns, we chose to stop pursuing this project. Since then, I have been checking out Google Reader blog with the hope to see Google start its recommendation service. Seeing that would have "verified" our critical thinking about the start-up. Today, I got the news, thru Google Reader of course. Here is the first pack of recommendations I got from 1000+ unread posts sitting in my account:

Based on my 3+ years in Google Reader, spanning 10,000s of read and starred items, I would expect a better "magic" from Google at the top ten bucket, but I am sure they will get there. With this feature, RSS became an even more important source of "learning", cutting down noise. With micro-sharing platforms plagued with 99.99% noise, self-promotions and non-value added posts, I actually feel good that millions of people are migrating from RSS to mediums like Twitter. That just puts me in front of the competition as they are spending hours and hours digging thru hundreds of 140-char posts to find 2-3 valuable items.