It's a great idea, and actually slashdot is already doing this on their front page. However, there are some problems: 1. If I want to find a particular item within the first "page" of results, but I've scrolled down 10 pages of results, find that item becomes pretty tricky! 2. Let's say I have a user who scrolls down 1000 pages. What does this to the browser? I'd imagine it grinds to a halt... The only solution would be to have a load page for the previous results. Here again, the issue I highlighted in point 1 becomes more emphasized. Especially if the search results change while it reloads (though you might keep the results for the user's session - but this sounds like a scaling issue).