Yeah, that part of Paul's talk was amazing. He said most people thought Gmail was a mistake, so he explicitly set a goal to make 100 users really happy, and just went round asking them "What exactly can I make this product do to make you happy?" Then he'd just pick things off the list and do them.
Expanding on the idea of 'redefining success', I love how he explained that if the answer to the "What can I do to make you happy?" question was too complex to implement, he'd just keep asking different people until he found problems that were relatively easy to solve.
