Harness engineering in practice in the age of AI

AI coding agents have made writing code fast, and many teams are now finding that what they ship this way breaks in production more often than what they wrote by hand. The agent usually works from a prompt and a repository, while a senior engineer works from years of accumulated context about the system, its boundaries and its failure modes. Teams that get lasting value from AI close that gap on purpose.

This one-day hands-on workshop takes you through the progression from prompt engineering to context engineering to harness engineering. We start with how the software development lifecycle changes once AI enters it and why authoring code works so differently now, then move on to building the guides and sensors that let an agent work inside your real architectural boundaries instead of guessing at them.

Who should attend

Senior developers, architects, and technical leads who already use AI coding agents in their teams, or are about to.

Topics covered

Module 1: What AI changes in software development

Why poorly used AI is damaging software that worked perfectly well until yesterday The four levels of code writing: developer-driven work, prompt engineering, context engineering, and harness and loop engineering

Module 2: Giving agents the context they cannot see

Capturing the human context that agents miss on their own, and using collaborative techniques to bring it to the surface Encoding the architecture directly for agents, so boundaries and conventions are explicit rather than assumed

Module 3: Building guides and sensors

Anticipating controls with fitness functions, testing architectural decisions as early as possible Building guardrails into the CI pipeline instead of relying only on manual review

Module 4: Harness and loop engineering in practice

The three loops that define how much autonomy to give an agent for a given task Which cycles can run without supervision and which need someone watching over them