There were impossible to beat spread sheets before Excel, impossible to beat word processor before Word, impossible to beat drawing programs before Illustrator (er... I don't know if that's actually the popular thing right now...) Those loosing applications were loosing customers well before feature parity in their rivals.
That's really the secret in competing against the established competition: don't build the parts of their app that suck, and spend a lot more time perfecting the parts that are great than the behemoth can justify. Work with the rifle like precision of a team of 10, rather than the shotgun like vagueness of a team of 1000. Know what's important, what's not and concentrate on execution.
The point he is making is not that you will be the next facebook, but that if you start with a small, possibly non scalable idea that works, who knows what it could turn into. And maybe its better to do this than to try to penetrate the market with big picture bullshit.