AI-assisted software development
- Audience
- Software developers and DevOps teams
- Prerequisites: Experience in software development, using Git and pull requests
- Platform: Claude Code by Anthropic
AI assistants now write code, tests and documentation, and carry out entire tasks independently. To ensure that the resulting code is something your team can understand and take responsibility for, rules are needed – and these are the very same rules that have always defined good software.
During the course, you’ll develop a small project entirely using Claude Code: from the initial requirements through to testing, the CI pipeline and pull requests, right through to deployment. I’ll demonstrate each step, and you’ll put it into practice yourself in the exercises: you’ll give the assistant tasks, check its work and specify how it should work. As with all my training courses, half the time is devoted to practical exercises.
TaskFest shows what this looks like in a larger project: a to-do list with a backend, a web frontend and a mobile frontend that I am developing with an AI assistant. The documentation and source code are on GitHub, along with the rules the assistant follows.
Contents:
- What AI assistants can do today, and where they fall short
- Setting up Claude Code: project instructions, permissions and hooks
- Describing tasks so that the assistant understands them correctly
- Rules for the assistant: conventions, asking questions rather than guessing
- Tests first: a failing test as a task for the AI
- The CI pipeline as a guard rail: build, tests, code analysis and security checks
- Pull requests and reviews for AI-written code
- Understanding, refactoring and documenting existing code using AI
- Data protection, source code and credentials: what the assistant is allowed to see
- AI in automation: Ansible, pipelines and infrastructure as code