Wired has an excellent article on how the Google search engine works and it's continued development.
What strikes me about this article is how many times the core Google search engine has been revised - 5 major changes to the fundamental search architecture over the last 13 years. Also, the details about how Google handles "bi-gram breakage" are awesome. This kind of processing requires an very high level of 'understanding' of the human world, and Google is doing it extremely efficiently. This is the kind of system that is going to wake up one morning and take over it's own development.