Basic research has a lot in common with startups. The unicorn rate is <1%, the best new ideas sound like bad ideas, and nearly all value from the best ideas is locked in the long-term future. The ideal startup/scientific program failure rate is not 0%, and could be 95%.

The current research system has serious problems, but we need accurate criticism to build a better future. "YC is all wasteful spending; why doesn't YC just choose to only fund the hits?" is absurd, but somehow we allow this argument when discussing NSF/NIH/DOE/DARPA.