By now, you've all been inundated with the news that Google is offering up a custom search engine, so I won't go into the details here and bore you to death. But my very first thought was, "hmmm....how can I get some SEO value out of this"?
What happens when you trim down a huge dataset into a much, much smaller one? You get a lot less clutter to wade through, right? Just seems like it might be a great opportunity to gain a little more insight into why one site might rank above another. Who knows, maybe there's nothing to be gained by it, but I think it's worth playing with to see if some ranking ideas might not surface. I'll let you know if I have a Eureka! moment from the testing.