All About Concave and Convex Agents — LessWrong