When it comes to the applications and the fear of missing good startups, I think the biggest bottleneck is how many groups they can interview.

If I'm not mistaken, for W2012 they interviewed 180 groups and accepted 65. That's 36% acceptance rate, but it shouldn't distribute evenly. If the applications were properly graded and the acceptance rate didn't get very low for the lowest graded applicants who were interviewed, then they need to increase the number of groups they interview. I guess that's the case, since pg previously commented that for S2012 they'd probably interview 270 groups. The "problem" is that for every new batch they probably get more applications (and most likely of higher quality), so this might be just playing catch up.

Maybe one day an Anybot robot will perform a preliminary interview so they can scale :)