I love this kind of idea, but it never works.

TechCrunch tested predictions of success - but what of failure? (maybe it says they all succeed...?)

Anyone care to enter a past learning-experience startup, to test this out?