CH Open - Open Source AI Workshops 2026

On 18 May 2026 I supported Joel Barmettler as co-host at the full-day workshop "The Vibe Coding Experiment" at CH Open's Open Source AI Workshops 2026. The event took place at the Berner Fachhochschule (BFH) and was part of a two-day workshop programme bringing together innovators, professionals and enthusiasts around open-source AI. Joel Barmettler, Senior AI Architect and MSc in Artificial Intelligence from Universität Zürich, had conceived the experiment and led the workshop. I supported him with the facilitation, team supervision and technical guidance.

The Idea: A RAG System in Six Hours

The central question of the workshop was provocatively simple: can a team build a working RAG system in six hours without a template, using only an open-source coding agent and an API specification? Four teams each took on one core component of the system: LLM inference server, vector database, document processor and chat frontend. Each team built its component from scratch, without ready-made frameworks, using only base libraries and opencode-ai as a copilot. At the end, everything was to be merged live into a working complete system.

Consistently Open Source

A deliberate choice of the workshop was the consistent use of open technologies. The only coding agent used was opencode-ai, an open-source tool that accompanied participants throughout the entire development process. The language models used were open-weights models such as DeepSeek or GLM-4, and the final RAG system also ran on a local open-source LLM. The workshop demonstrated that powerful AI systems can be built entirely with open technologies, without dependence on proprietary services.

The Format: Four Sessions, One Goal

The workshop was structured into four sequential sessions. In the first phase (Foundation), participants developed the system architecture together, received an introduction to opencode-ai and formed their teams. The second phase (Understand and Plan) gave each team time to understand their component in detail and prepare a short presentation. In the third phase (Build), the real work began: implementing the core functionality with opencode-ai as a copilot, followed by a progress check. The fourth and final phase (Integrate) brought all components together, culminated in a live demo and closed with a shared retrospective.

Who It Was For and What Participants Took Away

The workshop was aimed at anyone who wanted to understand how AI systems work under the hood and try AI-assisted programming with open-source tools. Programming experience was not a requirement; more important was the willingness to experiment. Participants only needed a laptop with a browser and a GitHub account; everything else ran in preconfigured GitHub Codespaces. They left with hands-on experience of Vibe Coding, an understanding of the core components of a RAG system and their own code to build on.

Conclusion

The experiment showed clearly how far open-source coding agents have come: within a single day, four independent teams produced an integrated RAG system. As co-host, it was rewarding to see how differently the teams approached their tasks and how the collaboration between humans and AI agents changes the development process. The workshop confirmed that Vibe Coding is not just a buzzword but a new way of working that functions with fully open tools as well.

Links

• CH Open: Open Source AI Workshops 2026, website.
• Joel Barmettler, main organiser of the workshop.