Good points. > I wonder what portion of the internet you have to scrape to be able to have good results for 90% of queries. But this is it. I don't think it'd

Good points. > I wonder what portion of the internet you have to scrape to be able to have good results for 90% of queries. But this is it. I don't think it'd be too hard to get 90% of queries right. It's the 10% that are the challenge. "Page rank" / "link juice" (amount/strength of incoming links) is still a main ingredient in Google's recipe, so they are actually able to determine, without manual intervention (more or less), which sites are "reputable", while your suggestion would require someone to curate the assortments of "reputable" sites. How would you discover niche blogs? Or even more high profile ones, just not known to the major public? That said, as you say, you could probably do with something like this for most of your day-to-day searches, and then just !g or whatever for the rest.