Human Agency at Stake — LessWrong