# Hypot

- **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:** [Faber Alberto Hoyos Ordosgoitia](https://cerebralvalley.ai/u/faberho)
- **GitHub:** https://github.com/faberho28/tutor-360
- **Demo video:** https://youtu.be/1QTpbaGfEPA
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2772

Tutor AI 360 is a web tutor that turns any camera connected to a PC, a webcam or even your phone used as one, into an expert looking over your shoulder while you repair electronics at home. No special equipment: a computer with a browser and a camera is enough.

The problem. Fixing hardware, whether a PC board, a phone or a keyboard, takes someone who knows where every part goes and what to test. Most people don't have that person nearby, and manuals or videos don't look at your device.

What it does. You open one page, point a camera at the device and say what you are working on. Then:
- SAM 3.1 outlines every part in the frame. No manual labels, nothing to train.
- Muse Spark names each part, judges its condition (ok, warning, anomaly) and gives a reason. That condition sets the color of the mask.
- Muse Spark writes a guided procedure for your goal: each step is a test ("if this happens, then that") on one numbered part, shown on the image beside it. After every answer (a tap, a note or your voice) it rewrites what remains and remembers the whole story.
- Each part has a diagnosis written from a close-up, which can say it cannot confirm a fault. Explanations adapt to four levels: junior, middle, senior and technician.
- Zoom with a contrast-enhancing magnifier and export a PDF report built in the browser.

It works for computers, phones, keyboards or any case you describe in a sentence. There is no device-specific code: Spark plans what to find from your description.

How it is built. Next.js, TypeScript and a Docker image. A light local tracker keeps the masks on the parts while things move; when movement is detected, SAM recalibrates and refines the masks.

## More from Meta Global AI Developer Hackathon

- [Doyle](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2769)
- [Diwan](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2770)
- [FlatSpot](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2771)
- [Tech4Good](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2773)
- [LambdaIans](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2774)
- [Cinemind](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2775)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2772. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt