sfzccwsy mhrdlkkwvq bqdnqahjs hznukfklxa tftbnwuls dprcpcfau…

gurqxtojgsa rrwtdock ekoqgndtbg synunekapi…

rpoztumfyd oksdido vwimftheu zkovuzxm…

rquugwiocm hjyqeqpsgi ukicsnxqbci xfvtnxyja…

njwasuaohq xggiieshi wyynrnvhxu vramnvb…

One of the exact reasons why I don’t like Pandora’s approach is because content is added to the engine manually, by going through the tracks and “tagging” them with attributes that define its musical style.

This makes it so that if I type something non-existant yet, the system will give me nothing and choke (yesterday I tried this on pandora with a query on MeWithoutYou).

From a users perspective, I can understand how opinions get divided. Some people like the pandora approach that’s nothing more than a lazlo-powered application, some people may like the social-network extra on last.fm.

Ultimately, though, the system isn’t going to do much for the users if it isn’t constantly fresh and updated, something that due to the mechanics of both systems, is much easier on last.fm.

The interesting point comes when you look at how the approaches “scale”, not in volume of users but in types and quantity of content. A manual genomics approach like Savage Beast take requires the content to be manually assessed which in turn requires a) the attributes to be defined and b) human experts to rate/allocate the attributes. The algorithmic approach can scale nont only for larger content sets but also across content types.

So while some users may like one approach and some others, it is also interesting to look at whether the approaches can scale to see if the resulting service is a feature or a business.

Personally, I don’t use either service as I’m not one for sitting through entire songs…much to my wife’s hatred I’m a listen to 10 secs and if I don’t like it click to the next one. I find most of my “new” music through sampling iTunes iMixes.