Workshop
Agentic Coding
Software development with AI agents: delegate, steer, verify. Your team works with coding agents from the first hour and leaves with a setup that survives the next tool update.
Prerequisite
Professional programming experience and everyday Git use. Any stack: TypeScript, Java, Python, legacy included. No prior experience with coding agents needed, healthy skepticism welcome.
Hands-on format
- Do first, then understand 🧪, every module starts with something you try, break, or watch fail.
- Live agent sessions 🖥, we steer real agents on real code, no slideware.
- Coding exercises 💻, built in your own setup and committed to your own repos.
What we cover
A menu of topics, depending on time, we pick the ones most relevant to your team together.
Inside the agent
- Build a working coding agent in a few hundred lines
- See that there is no magic inside
- The mental model for everything that follows
Context engineering
- Why more context makes results worse
- When to start a fresh session, what survives compaction
- Watch context rot live, then fix it
Steering
- CLAUDE.md / AGENTS.md, rules & skills
- Hooks & subagents
- MCP, and how it differs from a skill
- What loads when, and what it costs
Working in loops
- Plan mode & plan-then-execute splits
- Hand your plan to a fresh agent, see what you left implicit
- TDD with agents
Judgment
- Watch an agent game a badly defined goal, live
- Failure signatures of agent-written code
- When to read every line, when to spot-check, when to re-prompt
Costs & governance
- Tokens, pricing, and not burning money
- What data leaves the building
- Defensible configuration
Why with me
Since early 2026 I've built software almost entirely by orchestrating AI agents. That shift made review, testing and maintainability more important to me, not less. Before that: six years of hands-on training for engineering teams at banks, consultancies and DAX-listed companies.
More workshops
Jump straight to another workshop, no detour via the overview.
Bring this workshop to your team
Remote or on-site, tailored to your team's level.