Bigwill · Film a 20-second household cleanup on your phone; an AI agent judges it and turns it into robot training data
Film a 20-second household cleanup on your phone; an AI agent judges it and turns it into robot training data Robotics teams building household robots need footage of real people doing real chores, and it barely exists. Figure alone committed $1B and pays people to film housework. But 20–30% of collected footage is never usable: hands leave the frame, the task is half done, the light goes. Most collection apps delete the bad clips without telling anyone why. Buyers also now need a rights record for every clip, not a blanket claim over a catalogue, and most datasets can’t show one.
Film a 20-second cleanup on your phone and upload it. A SERV Reasoning agent checks it against 8 written rules in under a minute. Accepted clips get graded, paid in USDC and licensed to robotics teams. Refused clips get the exact reason, and every decision is published on a public ledger.
Figure’s in-house collection platform, data contractors running thousands of paid workers across 50+ countries, and closed filming apps. All of them judge clips privately, and the contributor and the buyer both have to trust a black box. STOW publishes every decision. Each clip is checked against 8 written rules with a written reason for each, accepted or refused, and code makes the call rather than the model. Every clip ships with its consent,…
35% contributor payouts in USDC for accepted clips · 25% product and engineering (new tasks, appeals, sybil protection, agent costs) · 20% staff salary · 20% marketing and robotics-team partnerships
No updates yet. The founder can post news here any time.
This project hasn't opened its fundraise yet. When it does, it runs on a bonding curve: anyone can buy or sell from the first second, and the price rises as the curve fills. Follow it to know when it goes live.
Every fundraise on Lifto runs on a bonding curve on Robinhood Chain. You can buy or sell at any time; each trade moves the price along the curve. When the curve fills, the token graduates to an open market pool. Early tokens are very risky: only trade what you can afford to lose.
Bigwill has 8 years in robotics across teaching, research and development. They taught robotics / embedded systems / AI at Carnegie Mellon University and worked on manipulation / perception / robot learning R&D. That work showed them first-hand that robots fail on everyday judgement, not hardware, which is why STOW focuses on demonstration quality instead of volume.