Should agents be durable?

Most AI agents never make it to production. The infrastructure underneath simply wasn't built for workloads that are unpredictable by design. In this workshop, Joey Baker walks through why traditional infrastructure breaks down under agentic workloads, what durability actually means at the task level, and how to express operational guarantees directly in your Python code. You’ll leave with a working production demo of self-orchestrating agents that don’t require any pre-provisioned infrastructure, custom retry logic, or separate config to maintain. Just annotated functions that handle parallel execution, survive failures, and give you full visibility into what happened and why. If you’ve ever watched an agent nail a demo and then spent months trying to make it reliable in staging, this one is for you.