Dunno if it works that well. I rate two or three sushi places (along with Chinese, Thai and American joints) and now my recommendations page has nothing but Japanese restaurants. In IR, this is called a diversity problem.
In the recommendation system space, it's called a sparsity problem (and/or a cold-start problem). You basically need to continuously feed it more data about your tastes for it to start doing a good job.
Sure, or it will go the way that Wave and so many other products before did. Or it will linger in zombie obscurity like Notebook did for so long. I'm not too worried about it either way.
