I will survey some results and insights regardings the cryptographic hardness of Constraint Satisfaction Problems, and relate them to questions in Computational Learning Theory.