Research and development of wireless sensor networks (WSNs) is motivated by their potential for ubiquitous monitoring and response . Expanded spatial and temporal monitoring coupled with actuation has huge promise for a range of defense, energy, agriculture, and manufacturing applications. However, WSNs still fall well below this potential. In this article I discuss reasons for this gap, based on lessons learned from developing multiple WSN systems. Specifically, I focus on the need for technologies to enable WSN software development, and rigorous models to improve the understanding of each WSN application.