Not a training. An event. Think hackathon, think LAN party, spread across several weeks with a group that wants to build for real.
The principle is simple: the group convenes, collectively picks a concrete problem, and builds the solution together. A product, a tool, a prototype — whatever the form, as long as it works at the end. The project combines artificial intelligence and, depending on what the group decides at framing, hardware components: sensors, connected objects, custom devices.
Example projects
- Specialized AI assistant. An assistant built for a specific sector, with a custom knowledge base.
- Automatic document analysis. A tool that extracts, classifies, and synthesizes information across large file collections.
- Connected object paired with an AI model. A physical device that talks to a language model to act on its environment.
- Agent-powered dashboard. An interface that aggregates and presents real-time data via autonomous agents.
The group decides. The group builds. The group delivers. The final result is published as open source, accessible to all.
What it requires
The format demands commitment. Each participant:
- Brings their own Claude Code or OpenAI Codex license
- Brings their own computing hardware
- Covers their own token, API, and credit costs
- Participates in the group agreement on hardware costs, where applicable
In exchange, everyone leaves with a shipped project, code in hand, and the experience of having built it alongside others.