Development and analysis of complex systems embedded in physical environments requires special modeling techniques. Development aids involve abstraction on multiple levels of representation in order to connect cross-cutting constraints and facilitate involvement of various stakeholders. Alternatively, run-time analysis requires type of modeling that is based on representation of interactions and behaviors. We discuss the Documentation Driven Development (DDD) approach for software development and Agent Based Systems for run-time analysis. We show that despite differences in architectures and functions both approaches possess important properties of reliability and flexibility and allow easy upgrading of their elements.