“Everyone understands that knowing your customer is key to success. But in the AI era, understanding how your LLM behaves is just as important. And it’s the part most teams neglect.”

This time, I’m excited to share something practical: an open source repo with the templates and playbooks I use to drive my AI-coding workflows:
👉 github.com/DamiMartinez/ai-dev-workflow-templates

This is a continuation (Part 2) of my previous post: My New Development Workflow - I Don’t Write Code Anymore. In this article I want to share how I manage references and context sources to help my coding agents (Cursor, Gemini CLI, etc.) stay accurate and up-to-date.

Over the past few months, my workflow as a developer has changed completely.
I don’t really “write code” anymore — instead, I define tasks, let AI coding agents handle the implementation, and focus on reviewing and improving the results.

Over the past months, I’ve been working a lot on developing AI agents and experimenting with different tools. Here’s an overview of the main AI tools and resources I currently use on a daily basis.

I have been working with the Google Agent Development Kit (ADK) for the past couple of months. While it is a powerful framework, the documentation is limited and I had to solve several unexpected problems on my own.