I will describe recent work on algorithms for ensuring generalization when random samples are reused to perform multiple analyses adaptively. I will also discuss connections to the problem of understanding generalization of algorithms for stochastic convex optimization and some challenging open problems.