E.NORRIS  The initial startup is ab it cpu intensive as we read in the information about the tracks you own and talk to the server. We are looking into it.  
(Version 2.4.3)

E.NORRIS  It is worth while trying The Filter more than once as it learns more about your collection through time, andby your usage patterns.

The recommendation algorithm is based on known relationships between songs, artists, or albums, genre etc. We return, by necessity, songs from your collection since we are not streaming music. We also do take into account your preferences, not just as you play a song ( i.e. skipping it will reduce it's chances of coming up, playing it and seeding it will increase the probablility) but would have read the iTunes database on startup and know your playcount, ratings etc.

The Filter will try and recommend the number of songs you request, you will find that if you have asked for a lot of recommenations from a band(s) for whom we are not overly familiar you may get "strange" recommendations but we will know some relationship between those songs based on people's playing, or purchase habits.

(The lack of familiarity can be because The Filter has not been run for long enough - most analysis of songs takes more time than the initial analysis stage, which is just a analysis of artists for speed purposes).

try again.  
(Version 2.4.1)