# Galileo – Physics from Video

- **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:** [keval doshi](https://cerebralvalley.ai/u/keval2504)
- **GitHub:** https://github.com/keval-doshi/galielo
- **Demo video:** https://drive.google.com/file/d/1eyrtHaNSqVoZhngGrI9qLmqyoNjDBP_R/view?usp=drive_link
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/800

Measuring physics usually needs lab equipment: photogates, motion sensors, or a stopwatch and many repeats. Most students and teachers only have a phone. Galileo turns a video of something moving into a physics measurement with an honest error bar.

You upload a clip, say what moves ("wooden block") and what you know (a length, a mass). SAM 3.1 segments and tracks the object in every frame. Muse Wonder then works like an experimental physicist: it reads the tracks and the frames, picks a physical model, fits it with real fitting tools, tests alternative laws against each other by BIC, checks the residuals, and writes a lab report with every step of its reasoning. Every number comes from the tools, not from the model's memory.

On a 5-second University of Toronto lab video of a rigid pendulum, given only "33 cm from the pivot to the centre of mass" and no calibration from the picture, Galileo measured g = 9.803 ± 0.089 m/s²: 0.04% from standard gravity and inside its error bar, from 70 seconds of analysis costing about $0.18.

It handles pendulums, springs, free fall, projectiles, ramps and domino chains, and can write and test its own laws for anything else. Each run gives an on-screen result and a downloadable PDF lab report.

## More from Meta Global AI Developer Hackathon

- [simTrain Universal](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/797)
- [Rikky](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/798)
- [ctrl+v](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/799)
- [GeneTrace](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/801)
- [NexusAG](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/802)
- [TerraSat](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/803)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/800. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt