Two-part workshop in German by Dennis Westermann 

It is possible to take part in the second part even if you have not taken part in Part 1.

Code Development with AI – Part 1: ‘The AI Working Environment for ILIAS Development’

14:00 to 15:30

No basics, no "What is ChatGPT": this workshop starts right where professional AI-assisted development stands in 2026. Live and without prepared slides, Dennis Westermann shows what a modern AI working environment for ILIAS development looks like: voice-driven prompting, an agent setup with an agent harness, subagents and an orchestrator, and why small, clean work steps are more reliable than one big leap.

The second focus is context and knowledge: vector databases and embeddings, an Obsidian vault served via MCP, and an AI-friendly knowledge base steer the AI towards modern ILIAS code and the right documentation instead of reproducing legacy patterns. As a running thread, a real ILIAS plugin is developed live in the background and carried on to quality assurance in Part 2.

All participants receive a companion repository with configurations, cheat sheets and links to rebuild everything at home. Time for questions is built in.

The workshop is held in German.

Code Development with AI – Part 2: ‘From Prototype to Quality’

16:00 to 17:30

Part 1's prototype is turned into dependable quality. After a short recap for everyone joining fresh (Part 1 is not a prerequisite), the focus is on a multi-agent setup with quality gates: evidence-based, positive testing of user stories and avoiding scope drift, with Dennis Westermann's open-source system "God Mode" as an example. The quality gates run visibly against the plugin developed live during the day; the goal is to see it running in a local ILIAS test instance at the end. A compact practice segment covers choosing MCP servers wisely, bundling tool calls into skills, and using hooks deliberately.

From 17:00, the workshop opens into a discussion round, explicitly welcoming non-technical participants and decision makers: What does large-scale AI adoption mean for ILIAS? Automated end-to-end testing, capacity questions, the "accountability gap", dependence on external model providers, and local LLMs as a safeguard.

All participants receive the companion repository.

The workshop is held in German.