x
Towards Sub-agent Dynamics and Conflict — LessWrong