The best apps are BETTER with critical mass, but work on their own. del.icio.us works just fine if you're the only user.
Twitter wasn't just on display. It served a real function. The audience WAS the conference goers. You can't really use it as a comparable example in this case. I seriously doubt a social network aimed at kids will achieve even a fraction of growth from TC50.
Whats important here is that the type of people that will pay attention enough to go to tc50/demo aren't just going to go there and then not say anything about it.
They'll come home, get on their macbooks and start blogging what they saw ( some really nice programmers building a kids network ). Some of their friends will check out the links, maybe share it on facebook, myspace, or , gosh, in an email. And so forth...
And for anything inherantly, viral-- an event like this is a big bump... It's compound interest with a larger initial principal. If you have a viral service that starts with 2 users and I have a viral service that starts with the 10k I'll get at TC50 and we both have the same viral co-efficient, who wins?
And for any company that is fundraising, the visibility/credibility provided by this event is HUGE.
And for any company that is actively looking for partners/bizdev deals, the visibility/credibility provided by this event is big.
Is the value big enough to be worth shifting your launch date dramatically? Probably not. But if your launch date is in the neighborhood, trying to get it done at TC50 (or a similar event) is a low-cost/high-value proposition. Hell, even DEMO is low cost compared to most other PR/marketing spending you could do.