x
AGI safety from first principles — LessWrong