Agent-based modeling—in which computers model complex systems from interdependent "agents" may be complexity theory's most promising tool. Its full potential hasn't yet been realized, but this lecture offers a taste of what it can already achieve in disciplines as disparate as fire prevention and disease transmission.