Great interview that reinforced some of PG's ideas that I am familiar with from his essays. The one that intrigues me the most though is that all really great ideas look like crazy ideas to begin with. While very counter-intuitive this seems to be true and is borne out by the data from actually funding hundreds of startups. [1]

So here is a question for PG. At one point in the interview he says that the majority of applications that YC gets seem to be terrible ideas.

"And so I have seen huge numbers of terrible ideas, you know? No one has seen more terrible startup ideas in the world, I'm sure."

If [1] is true, wouldn't that imply that this set of terrible ideas, may in fact contain a few gems - some really great ideas? In fact, based on the analysis and data, one could even argue that the magnitude of the effect (how good the idea is) may be proportional to how bad it sounds in the beginning! So in selecting startups to fund, how does YC distinguish between terrible ideas that would actually be terrible, vs terrible ideas that would actually turn out to be great. Maybe there is no simple way except a gut feeling, or based on how good the founders are. Or maybe the only way to really find out is to actually build the product and let the market decide?

[1] http://www.paulgraham.com/swan.html