Agentic Playbook

A learning site that documents the spec-driven, memory-backed way I build with coding agents. Built in public for engineers who want a real method, not vibes.

Most "build with AI" content is vibes. People share screenshots of one good Cursor session and call it a method. That doesn't help anyone ship the second or third project. The actual hard parts — keeping context across sessions, scoping work into the right size for an agent, avoiding rework — never make it into the tutorial.

Agentic Playbook documents the way I actually build with coding agents: a seven-step flow that goes from brief to shipped, with persistent project memory at the center. Each step has an explicit purpose — research, scaffolding, TAG-style scoping, debug protocols, review, ship — and each one is a lesson I wrote because I needed it on my own work. The site itself is built in Astro with MDX so the lessons can include real prompts, real diffs, and real screenshots without fighting a CMS. I update lessons as my own practice evolves; nothing on the site is frozen tutorial fiction. The point is not to sell a course. The point is to be honest about a method that works for me and let other engineers steal the parts that fit. If you've read three lessons and shipped one project differently because of them, the playbook did its job.

Agentic Playbook is live, indexed, and accumulating real lessons drawn from real projects. It is the proof artifact behind the rest of the portfolio: the same method built 1mins.in, FrameStory, and the router. Shipped, indexed, and finding its first real users.

• Seven-step build flow with persistent project memory
• Each lesson is something I actually do, not a tutorial cliché
• Proof I learn by going deep and teach by doing