Multihop and multiflow paradigms are expected to play a central role in future wireless networks, enabling spatial spectrum reuse and a transition to heterogeneous network scenarios. This raises exciting new information-theoretic questions. In this talk, we focus on three such questions. We first ask “how does network topology impact information flow?” and show how the DoF of two-unicast networks can be characterized based on the network graph. We then study “to what extent can multihopping improve interference management?” and present a technique that fully diagonalizes two-hop networks. Finally, we ask “how robust are network models to nonidealities?” We describe our characterization of the Gaussian noise as the worst-case noise in networks and how this yields new capacity outer bounds.