UnSearch

LessWrong

Understanding Mesa-optimization Using Toy Models

LessWrong post detailing and motivating our Research Agenda.