Welcome

I have been coding with AI agents almost every day for a year now. Today, I delegate about 99% of my coding work to Claude Code. I’ve tried many different practices along the way — some worked great, some didn’t work at all.

This playbook is where I share what actually stuck. The patterns I use every day, how my workflow evolved over time, and the things I wish someone had told me earlier.

Three levels of AI coding

The way I see it, there are three levels in how people work with coding agents.

Level 1 — Prompting. This is where everyone starts. You drop a prompt, expect the result to be right. Sometimes it is, sometimes it’s not. Most people stay here and get frustrated.

Level 2 — System. You start building patterns around the agent. Context management, delegation to sub-agents, AI reviewers, hooks. You create a system that produces consistent results instead of hoping for them. This is what most of this playbook covers.

Level 3 — Framework. Multiple reviewers, specialized skills, orchestration layers that deliver high-quality code consistently. This is advanced territory.

This playbook is for people who want to get started and progress to Level 2. If you’re already comfortable with context management, AI reviewers, hooks, and delegation to specialized agents — you might be ready for Level 3. Look into repositories like Bryan Finster’s Agentic Dev Team, nWave, or Claude Superpowers. I follow these people, read their code guidelines, and adopt a lot of their practices into my own workflow.

I’m not a fan of heavy frameworks myself. I prefer spending more time doing things step by step — understanding every detail and fine-tuning everything for how I work. That’s my preference, not a recommendation. These frameworks are good. They’re just not for me.

A bit about me

I’m a software engineer and consultant. Beyond my own projects, I coach teams on AI coding adoption — helping them go from “trying Copilot sometimes” to building real features with agents. This playbook is essentially the same material I use with my clients, just written down.

I’m not claiming to have all the answers. AI coding is moving fast and we’re all figuring it out. But after a year of daily practice, I’ve developed habits and patterns that consistently produce good results. That’s what I want to share here.

What’s in this playbook

• Core Concepts — How I manage context windows, structure my sessions, and prompt effectively. The fundamentals that make everything else work.
• Feedback Loop — The guardrails I set up for every project — builds, linters, code review agents, tests. This is what keeps AI-generated code on the rails.
• Orchestration — How I coordinate multiple agents, split work across swarms, and think about team structure for AI-assisted development.

Each page covers one topic with practical examples you can apply right away. Read in order or jump to whatever interests you.

Your feedback is welcome

I want these articles to be easy to read and genuinely useful for your productivity. If you have ideas for topics, improvements, or just want to share your experience — drop me a message. I’d love to hear from you.