Before the event
- Share the workshop URL and ask participants to bring an AI coding agent that can fetch URLs.
- Decide whether deployment is required or optional. Sandbox SDK and Workers for Platforms access may gate capstone milestones.
- Prepare a shared channel for questions and one checkpoint slide with the three-plane model.
- For larger groups, pre-select one capstone path so mentors can support common issues.
Opening script
- Explain that the workshop teaches platform primitives, not just how to use a vibe-coding tool.
- Ask participants to open the homepage, enroll, and copy the prompt into their AI agent.
- Tell them their agent will fetch
/llms.txt, guide each lesson, quiz when appropriate, and mark progress complete. - Set the first checkpoint: everyone should be able to explain control, execution, and deployment planes.
Live checkpoints
- Architecture: participants map a user prompt through control, execution, and deployment.
- ProjectAgent: participants explain why project-scoped state beats user-scoped state.
- Sandbox: participants identify which code is untrusted and where it runs.
- Preview: participants distinguish a dev-server preview from production deploy.
- Capstone: participants choose Tiny Platform, Code Review Agent, or Sandboxed REPL.
Facilitation tips
- Let the AI agent personalize explanations. Keep facilitator talk short and focused on room-level decisions.
- When participants hit account blockers, switch them to design-review mode rather than stopping the group.
- Ask participants to paste one command output, architecture decision, or acceptance-check result into the shared channel.
- For mixed-experience groups, pair advanced participants with beginners during the Sandbox and Workflow checkpoints.
Closing prompt
Ask each participant to write down the boundary they understand better now, the Cloudflare primitive they want to try next, and the capstone milestone they will continue asynchronously.