AI agents can generate code faster than we can understand it. That gap creates verification debt: changes reach production before anyone has validated what they do.
In this talk, you will learn the Context Stack, a layered model for how developer intent becomes working software in agentic workflows. You will also learn a shared responsibility line that clarifies what is on the model or platform versus what is on you.
We will use the stack as a diagnostic tool to identify whether failures come from missing project truth, lost session state, unclear constraints, or weak verification. Then we will translate it into practice with language-agnostic patterns: spec-driven development (requirements to design to tasks to verification), context loading strategies that avoid blowing up the window, and lightweight quality gates (tests, lint, security checks) that keep ownership where it belongs. With you, the developer.
Banjo is a Senior Solutions Architect, where he helps builders ship with generative AI. He focuses on agentic workflows, developer experience, and the practical craft of building reliable AI-powered systems. Outside of work, he runs a local meetup, builds open-source agent tooling, plays JRPGs, and explores whatever events are happening nearby....