Shared publicly  - 
Illustrating how hard it is to create a good recommendation engine, Apple, Amazon, and Google every day shower me with exciting new offers almost universally uninteresting to me. (At least when it comes to music and video; Amazon is a bit closer to home with products, which apparently is easier since it often just regurgitates what I've searched for.) For music, the most consistently reliable recommendation engine is direct communication with my friends. This 1956 recording of Nathan Milstein performing Bach's sonatas and partitas is one example.

Generally I don't find "I'm listening to XYZ" status messages on social networks very compelling. But in case you do, here's my little contribution. I hadn't paid close attention to this on Google+, but apparently if I've put you in one of my circles, you get a full listen to this; if not, you only get a sample.
LeGrand Johnson's profile photoStephen Shankland's profile photoGordon Haff's profile photo
This set is the first recording of these pieces I ever heard. My father listened to them when I was little. The performance is excellent, but these are my favorite recordings of the Sonatas and Partitas for nostalgic reasons.
+LeGrand Johnson I'd listened to the later Milstein recordings (1975?) but a friend whose taste I really respect recommended these from 1956.
The main (well, one of the main) issues I find with Amazon recommendations is that it doesn't seem to try to understand the relationship between product purchases. In other words, if I buy product X it might mean I'd be interested in Y and Z (if it were a book on a similar theme, say) or it might mean that I just bought a big screen TV and therefore I'm really quite unlikely to buy a similar TV in the near future. Makes more sense if I've just been browsing of course. 

In general, even bounded product recommendations such as music, movies, or books is pretty hard for a variety of reasons. Expand it to the general case and it seems to be pretty much impossible.
Add a comment...