1.) Reddit. The day after it launched: "Well, this is interesting, but I can't ever see myself using it." 3 months later, after I started using it: "Well, I use it, but I can't ever see it getting popular." Early 2006, in response to PG's comment about "This sucker's profitable": "But are they profitable enough to cover the opportunity cost's of the founder's salaries?" Mid 2006: "Okay, they're profitable through bizdev deals, but I can't imagine them ever getting bought." I believe they were acquired on Halloween 2006.
2.) Infogami: I thought they were the most likely of the SFP05 startups to be successful, right until they launched.
3.) Wufoo, when it came out: "I don't get what the point is." About 9 months later: "Incidentally, I'd put Wufoo, VirtualMin, and DropBox at the bottom of the deadpool, i.e. most likely to succeed."
4.) Xobni: "Personally, I'd put Xobni at the top of the YC deadpool." http://news.ycombinator.com/item?id=51982
5.) VirtualMin: Already mentioned that this is one I'd invest in if I actually had money. I'd used Webmin for previous websites; there's a lot of money in easy-to-use server admin tools for folks who don't really know what they're doing.
6.) RescueTime: Loved it from the moment I heard about it. Immediately recommended it to friends; if I had money, I'd invest.
7.) Thinkature: I thought this was really cool when it came out, and that it'd be one of the survivors of its batch.
8.) Flagr: Thought this was succeeding for sure when one of my RL friends linked me to a Flagr map.
9.) Justin.TV. "This is pointless, I can't believe YC gave these guys $50K, I'll probably never visit it again."
Somebody remind me never to become a VC.