# Playspec

- **Event:** [Gemini 4 Argon SF Hackathon](https://cerebralvalley.ai/e/gemini-4-argon-sf)
- **When:** Sun, Oct 25 at 9:00 AM – 10:00 PM (PDT)
- **Where:** San Francisco, CA
- **Team:** [Robert Amanfu](https://cerebralvalley.ai/u/ramanfu)
- **GitHub:** https://github.com/retroam/argon-craft
- **Demo video:** https://youtu.be/gGw4fKxgAC8
- **Gallery:** https://cerebralvalley.ai/e/gemini-4-argon-sf/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/gemini-4-argon-sf/hackathon/gallery/37

Hardly anyone can write a full game spec before they've played anything, so this project is an attempt to simplify the game creation process. You sketch a level on a 2D canvas and play three 20-second 3D prototypes built from it. Gemini 4 Argon then works out what you want from how you played.

While you play, the app records how you move and that telemetry goes to gemini-4-argon at thinking_level: "high". The call is chained to the earlier canvas and prototype calls with previous_interaction_id, so Argon sees the whole session. It returns a spec (your player type plus tuned physics) and generates the full game as a single Three.js page.

Then the loop runs. Playspec renders the game, captures a WebGL frame, and sends Argon the frame, the scene graph, and your telemetry. Argon critiques the scene against the spec it inferred. The app patches the level, rebuilds the game, and runs it again. Each pass shows the frame Argon saw, its critique, and a diff of what changed.

You can export the finished level as a Unity scene file.

## More from Gemini 4 Argon SF Hackathon

- [Team Real2Play](https://cerebralvalley.ai/e/gemini-4-argon-sf/hackathon/gallery/34)
- [AI Pitch Studio](https://cerebralvalley.ai/e/gemini-4-argon-sf/hackathon/gallery/35)
- [Get Me An Alt](https://cerebralvalley.ai/e/gemini-4-argon-sf/hackathon/gallery/36)
- [Solo - Care in Concert.](https://cerebralvalley.ai/e/gemini-4-argon-sf/hackathon/gallery/38)
- [Heavy Rain](https://cerebralvalley.ai/e/gemini-4-argon-sf/hackathon/gallery/39)
- [Storm simulation agi](https://cerebralvalley.ai/e/gemini-4-argon-sf/hackathon/gallery/40)

---

Markdown version of https://cerebralvalley.ai/e/gemini-4-argon-sf/hackathon/gallery/37. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt