In late 2001, my research group initiated a project for the US Army to explore the integration of intelligent agents, security protocols, and mobile ad-hoc wireless networks (MANETs). The resulting Secure Wireless Agent Testbed became the basis for over a dozen follow-on projects that theoretically or empirically (often both) examined phenomena that appear when building decision making algorithms in environments where communications is not guaranteed or free. Where prior work assumed the communications fabric to be ubiquitous and perfect, our research explored real-world issues that create havoc for traditional distributed AI algorithms and developed techniques to mitigate these issues. The technical insights we produced were highly multi-disciplinary, including numerous publications in both AI and networking venues. This talk will introduce some of the key ideas from this work and present two of them in detail: distributed constraint reasoning and the development semantics for content-based networks.