You know AI can change how you code. You just haven't found the right method. A self-paced course built from months of production practice — or hands-on consulting for your team.
By Jean Desauw — Lead React Native Engineer at Odisei Play, building agentic workflows in production daily.
You tried Claude Code for an afternoon, found it mediocre, and gave up. That's normal — it takes structure and practice.
When it works, activation energy drops 80%. Starting a new module, refactoring an entire layer — what used to take a full day of motivation takes an hour of setup.
A good prompt isn't enough. Your repo structure determines 80% of the agent's effectiveness. The shift isn't technical — it's methodological.
You read blog posts — tips, no structure. You read tool docs — features, not workflows. You looked at “AI consultants” — slides, no production experience.
Built from months of agentic coding in production at Odisei. Not theory — methods that work on real codebases.
One system, meaningful checkpoints. You spec, the agent builds, you review where it matters. The level of autonomy adapts to the task.
The repo structure determines 80% of the agent's effectiveness. CLAUDE.md, conventions, and organized code — before your first prompt.
Skills loaded on demand, MCPs for live tools. The agent works better when it gets what it needs — not everything at once.
Every friction point becomes a system improvement. Capture what breaks as it happens. The method evolves constantly.
Six months from now, half of this might be outdated. That's why the method matters more than any specific setup.
Rich text lessons with code examples, callouts, and exercises. Built for reading — not watching.
What to include, what to exclude, and why this single file determines everything that happens next.
When Claude Code enters a project, the first thing it reads is CLAUDE.md. This is the single file that determines whether the agent understands your project — or guesses.
CLAUDE.md is not documentation. It's an operational briefing. Write it for someone who will start working in 30 seconds.
# CLAUDE.md
## Commands
pnpm dev # Start dev server
pnpm build # Production build
pnpm lint # ESLint
## Anti-Patterns
- Never import @jd/db/server in client
- Never call AI providers directlyThis is what a lesson looks like. Rich text, code blocks, callouts, and exercises — optimized for learning, not consumption.
Learn at your own pace or get hands-on help for your team.
For individual developers
A self-paced course that takes you from understanding how Claude Code works to building your own production-ready agentic system. What took me months of practice, structured so you can start in hours.
Early bird — first 10 buyers only. One-time payment, lifetime access
For teams & companies
Your team uses AI tools individually with no shared method. No common workflow, no clear ROI. I did this shift for a real team at Odisei — I can help yours.
From the fundamentals to a complete, working blueprint. Each chapter has exercises — this is not passive content.
One system with meaningful checkpoints. Spec writing, plan mode, course-correcting, and the identity shift.
Skills vs MCPs, subagents, and the decision framework. Give the agent what it needs — not everything at once.
Capture friction as it happens. Hooks, evolving your CLAUDE.md, and giving Claude a way to check its work.
Parallel sessions with git worktrees, non-interactive mode, the writer/reviewer pattern, and permissions.
The deliverable. Audit your setup, write your CLAUDE.md from scratch, create your skill library. Walk away with a system that works.
Chapter 1 is entirely free. No account needed. See if the approach clicks before you commit.
Your developers use AI tools individually. There's no shared method, no common conventions, no way to measure what's working. I help your team build that system.
Look at your current setup — codebase, tools, habits. Understand where you are.
Configure the tools: Claude Code, Cursor, Windsurf. CLAUDE.md, repo structure, conventions.
Hands-on session with your team. Adapt the method to your context. No one-size-fits-all.
A document the AI agent uses directly. I leave, the method stays. Your team keeps improving.
You can — it took me months. The course compresses that into a structured path with exercises and a practical deliverable. The first chapter is free — see if it's worth your time.
Some of it will. That's exactly why the course teaches the method, not just the tools. The principles — codebase-first, context management, feedback loops — transfer even when the tools change. I'll update the content as things evolve.
The course focuses on Claude Code because that's what I use daily in production. But the principles — context management, spec writing, feedback loops, codebase structure — apply to Cursor, Windsurf, and any agentic coding tool.
Yes. Chapter 1 starts from the fundamentals. I was a junior without a mentor not long ago — the course is built so you don't need one either. The free chapter will tell you if it's the right level.
Then you'll skip the basics and go straight to context budget management, feedback loops, and scaling patterns. Most daily users haven't optimized these — that's where the biggest gains are.
The entire first chapter is free. Try it, see the quality, then decide. That's more honest than a 30-day refund policy. If the free chapter doesn't convince you, don't buy.
I also do consulting for teams — audit, workshop, and a blueprint.md tailored to your setup. Scroll down or reach out directly.
The course is in the works. Join the waitlist and lock in the early bird price — €49 instead of €90. Limited to the first 10 buyers.