Advances in quantum computing hardware now enable running substantially large experiments, exploring both NISQ applications and demonstrations of quantum error correction (QEC). In this talk, I will present two applications of tensor networks we have recently used within these two contexts. First, I will discuss improvements to a powerful and widely used adversarial method employed to challenge the beyond-classical nature of certain NISQ experiments. This approach involves optimizing tensor network contraction schemes subject to realistic memory and performance constraints imposed by current supercomputers. Second, I will talk about our general-purpose maximum-likelihood tensor network decoder. We have used this optimal decoder to benchmark our recent QEC experiments, achieving the first experimental demonstration of error suppression on a surface code. In addition, we have leveraged this decoder to evaluate the performance of alternative, more scalable decoding strategies.