Arbital project outline — LessWrong