Skip to Main Content

Proof of Turnover

Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

Proof of Turnover — Demo video

I have vacation rentals in Charleston, SC. Between two guests my cleaner has to reset around 40 different things in the house, and all I get is a few random pictures in WhatsApp. I can't really tell if there are fresh towels, a new toilet paper roll, or a pile of laundry left on the floor. I just hope the house is ready. Airbnb alone has over 9 million listings, so every host has this problem after every checkout. So I built Proof of Turnover. The cleaner opens a link on her phone (no app to install) and the app walks her room by room, one photo at a time, in Spanish, English or Russian, and reads every instruction out loud. While she keeps going, every photo gets checked in the background. Something only counts as done if Muse Spark says it's done on a specific photo, a second separate Muse Spark check agrees on that same photo, and SAM 3.1 actually finds the thing there (and enough of them, like 2 towels). If something is wrong or not clear, she gets a card with a red box on the photo and one short instruction. She fixes it, takes a new photo, and only that card gets checked again. If she says "it's fine", Muse Wonder gives a second opinion. At the end I get ready or not ready for the next guest, with a photo for every item and before/after for what she fixed. You can try it without an API key. There's a recorded real turnover in the repo, see the README.

Team