Session: How the Oldest Principle in Engineering Will Reshape AI
A fundamental principle in engineering—the decomposition of complex systems into modules with well-defined interfaces—has long enabled the transition from prototypes to production. Generative AI has yet to fully make this shift. Today’s agentic systems often remain monolithic and fragile, embedding logic, evaluation, and recovery into a single end-to-end prompt.
We propose treating both AI models and applications as structured, extensible systems rather than monolithic artifacts. By adopting composable architectures, capabilities can be added, tested, and evolved independently. As modularity replaces monolithic design in AI, it will become the foundation for greater reliability, scalability, and collaborative development.
