Paul Crowley comments on MathOverflow as an example for LessWrong