I've written before that EA can be viewed as a system. There are certain roles that, if filled well, cause the system as a whole to be more successful. For example, 80,000 hours ensures that EAs end up in jobs where they can have an impact and university EA groups ensure a fresh supply of top graduates. What role, which is currently unfulfilled or not fulfilled well is the greatest bottleneck on the capacity of EA to succeed?
A larger number of formidable people, in Paul Graham's sense of the word.
Here's some related recent discussion on LessWrong.
I have a bunch of thoughts about this that'd be easier to convey elsewhere. Shoot me an email? (here)