Compounding Incumbent Advantage — LessWrong