Superintelligence 29: Crunch time — LessWrong