Makes sense. Obviously books are an area where search alone won't cut it. Knowing that a book exists and knowing you would like to read it are very different questions. How do you identify the boundary between search and trusted curators? How do you integrate trusted curators with collaborative filtering? How do you filter out books filled with rookie mistakes?

A similar issue is being worked out in news. Most people would still rather read the NY Times and WSj rather than wikinews. The reason: editorial input. Wikinews is a noble effort but has not got the magic formula for editorial control.