[startup critique] Randomicity - Random sites from across the web

This is a little side-project that I've been working on for about a week, in between my main startup & taxes & moping around from being rejected by YC. It crawls the web for sites (only frontpages, not the full text like Google), and then displays a random selection of 20 on each refresh. The problem I'm trying to solve is that I'll likely need a new idea for my startup, yet there's a whole World Wide Web out there and I've never explored 99.9999% of it. Most social news sites try very hard to give you sites you'll like, but I don't want sites that I like; I want sites I've never heard of.

I initially had the idea for this about a year ago, but dismissed it as "Something so simple couldn't possibly be useful." But recently, I got to thinking about Paul Buchheit's blog posts on humility, and Linus Torvald's advice on starting from a trivial problem and expanding it, and Evan Williams' success stories with Blogger and Twitter. Maybe humility isn't just acknowledging that the big idea you've pinned your hopes on may fail; maybe it also involves acknowledging that all the little ideas that you've dismissed as being incredibly stupid might also succeed. So here's my incredibly stupid little idea. If it fails, I only invested 3-4 coding days in it, and it'd be worth it just for the Django and BeautifulSoup practice.

Unlike past side projects, I will be maintaining and updating this, as long as it remains fun (i.e. as long as there're cool hacks that'd make it noticeably more useful). I don't have any grand plan for it; my development process is basically "Look at feedback e-mails, implement those, look at the site, add any features that I think would be useful." If you've got ideas, submit it through the feedback form (which I want to replace at some point; I'm using James Bennett's django-contact-form plugin, and it's a bit heavyweight for this purpose) or post here.