# TEARSHEET M

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Malik Zhang](https://cerebralvalley.ai/u/MalikZ)
- **GitHub:** https://github.com/Malik1942/tearsheet
- **Demo video:** https://youtu.be/_TyiYuu7dOM
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4737

Tearsheet turns the photos you already take into design rules, then lays out product screens from them, with every choice traced back to a spot on a photo. Taste you can trace.

Live: https://tearsheet.malikzhang.com
Try it: open the link, click Make, type what you're making, and press Enter. The library opens with sample photos, so you need nothing of your own.

The problem: we save photos because of how they look, like a stage at a show or a car in a garage. When a new project starts, none of it makes it into the design. A moodboard holds pictures but doesn't say what to do with them. A chat can give advice, but you have to dig the photos out and describe your taste every time, and its answer is a one-off you can't adjust piece by piece.

How it works: Muse Spark does two jobs. When you save a photo, it reads it once and keeps up to three rules, each tied to a measurement with a box on the photo. When you say what you're making, it picks the photos that fit by reading their rules, never the photos, and quotes the rule behind each pick word for word. A fixed layout shell then builds two real screens from those rules; no model draws them. Each photo is a block: turn it off or swap it, and only what it carries changes. Every measured value points back to the box it was read from. When you're done, share one read-only link, or export tokens and a handoff file for a coding agent.

It's for product designers and indie developers who want their apps to look like their own taste, not a template. That's Build for Your World taken literally: the library is your own camera roll.

Before the event: only my photos and the credited sample images existed. All code was written during the hackathon, Oct 26 to Nov 4, 2026.

## More from Meta Global AI Developer Hackathon

- [Critics wonder](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4734)
- [SkyNet](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4735)
- [George Sung](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4736)
- [Three Dimensions](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4738)
- [Project Prometheus](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4739)
- [QQS Interactive (solo)](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4740)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4737. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt