The Ubiquitous Converse Lawvere Problem — LessWrong