In this issue: we discuss what to test in ML models; we explain how Meta uses A/B testing to improve Facebook's newsfeed algorithm; we explore Meta's Ax, a framework for A/B testing in PyTorch.