AI-assisted projects rarely fail because of the model. They fail because of missing specifications, unreviewed results, and configuration management that nobody can trace anymore. That's your home turf as an assessor – just approached from the other direction: you know the process. In this training, you run it yourself.
No slide deck and no certificate of attendance – instead, a repository you'll keep using afterward.
Every training day ends with a tangible state in the repository. On the last day, there's a release – and the path there is fully traceable.
How a coding agent actually works: context, limits, typical failure patterns. Installation and terminal basics for everyone, prompting for development work – and in the afternoon, you present your project idea and lock it in.
CLAUDE.md as a project handbook for humans and agents alike. Requirements with acceptance criteria instead of wish lists, designing architecture in Plan Mode before any code is written – and recording decisions with rationale as ADRs.
Explore – Plan – Code – Verify as a controlled loop. Tests first, reading diffs, stopping wrong turns early. Debugging, subagents, and parallel work streams – by the end of the day, the first walking skeleton runs.
Your own workflows as slash commands, hooks, and skills. Connecting MCP and external interfaces, managing keys cleanly. Data privacy, permissions, and licensing – and then your project goes onto a real server.
Close out and document open items. Building evidence in the repository: traceability and review records. Then Demo Day – everyone presents their own result. To close, you put your assessor hat back on and evaluate AI-assisted development.
To take with you: The full training outline as a PDF – 28 pages covering the agenda, learning objectives, detailed daily schedules, prerequisites, and outcomes for each day.
Download PDF (40 KB)Budget for €50–80 in incidental costs of your own. All accounts are set up in your name and remain in your hands.
Five full training days spread over two weeks – with at least one free day between sessions. Time to keep working on your own project, gather questions, and let what you've learned sink in.
This first delivery is aimed mainly at friends, acquaintances, and long-time colleagues – which is why, alongside the discounts, there's also the option to pay with time instead of money.
Time for time instead of an invoice: you take part in the full five days and, in return, contribute two and a half days of effort to a SPICEy3 project. What exactly, we'll work out together beforehand – it should match what you're already good at.
Request a Quote