Subagents of Cartesian Frames — LessWrong