Our coding agents are getting smarter, but the workspace we drop them into is probably not enabling them to use their full potential. The decisions, failed attempts, conventions, and product knowledge are still crucial context scattered across people’s heads, Confluence, old conversations, and forgotten pull‑request comments.

During this talk, we'll walk through the shift from the “good old” prompt engineering to agent experience, then I'll show a practical AX experiment at my current company that led to a better workspace for both agents and the people using them.

Coding agents can write a lot of code, but they sometimes end up in a dead end and learn the toolchain by trial and error. How can you ship with more confidence when agents write your code? We'll look at how Expo moves knowledge from the docs into tools designed for agents: the Expo agent CLI, co-located skills, and EAS Simulators. Together, they let an agent run the app, test it, and give you confidence that your goal was achieved.

Augmented Coding is more than just asking an AI to help you write code.

You externalize your decision points, heuristics, and workflows into artifacts the agent can follow. The agent becomes your mentee.

As you do that, you change your focus from low-level details to high-level intent and methodology. Decisions that were unconscious become explicit.

In this talk, I’ll share my story of how I learned to teach AI a Software Crafter’s approach to development.

The agent follows a rigorous TDD process with 121 unique nodes and 131 directed edges that can run autonomously for hours, creating dozens of commits.

It covers advanced context engineering and techniques to increase autonomy while aiming to maintain quality.

And I will tell real stories of AI-enabled breakthroughs in real-world product development.

I know, I know - another AI talk. But hear me out: this one isn't about prompts, it isn't about replacing developers, and ChatGPT won't be mentioned once. It's about what happens when you stop using AI as a fancy autocomplete and start wiring it up with tools, memory, and the ability to orchestrate other agents. We'll look at how Claude skills and plugins actually work, build a small agent pipeline live, and be honest about where it breaks. You might leave with ideas you can actually use on day-to-day bases.