x
On morality, defection-robustness, and legibility incentives — LessWrong