Many of us are experimenting with AI coding assistants—but how do you turn “neat tricks” into a repeatable workflow your whole team can use?
This session shares a practical, repo-native pattern we use at Stanford’s Engineering Center for Global & Online Education (CGOE) to keep AI-assisted development fast, safe, and easy to hand off. Instead of one-off chats and copy-paste, we’ll walk through a lightweight structure built on four building blocks:
- Timestamped session notes tied to commit IDs, so any engineer (or AI) can pick up where the last one left off
- A simple TODO system in Git for scoping work and tracking progress
- Docs-as-code guides under
docs/guides/so fixes and runbooks live next to the code - A testing strategy that emphasizes integration and end-to-end checks to keep changes small and safe
We’ll also show an “AI arcade” visual demo—classic 2D games built in minutes using CLI assistants, tests, and this workflow—so you leave with both prompts and concrete examples you can adapt the same day.
