1 hour ago
9
Figure AI just turned your morning coffee routine into training data for robots. The company launched Index, a crowdsourced platform that pays people to record everyday tasks on camera, then feeds that footage into the AI models powering its humanoid machines.
The app emerged from four months of stealth development under the codename Project Go-Big. By its public launch on August 25, Index had already racked up 264,000 downloads across 108 countries and more than 44,000 weekly active users.
The world’s largest robot training dataset
The numbers behind Index are staggering. Contributors, whom Figure calls “Creators,” have uploaded more than 16 million videos to the platform. The system processes 30 minutes of video every second, which translates to roughly 4.9 years of human activity uploaded per day.
The diversity metrics tell an interesting story. Per 1,000 hours of collected data, the platform averages 373 unique tasks, 1,146 unique manipulated objects, and 116 unique environments.
Figure AI has paid $15 million to contributors since the platform’s inception.
From video to robot brain
Raw video alone doesn’t teach a robot anything. Figure runs all incoming footage through a five-stage curation pipeline before it becomes useful training material. The processed data feeds into Helix, the company’s vision-language-action AI model that controls its Figure 03 humanoid robots.
The motivation for building Index apparently came from frustration with third-party data vendors, who couldn’t deliver the diversity and quality Figure needed.
A billion-dollar bet on scale
Figure AI has committed more than $1 billion to scale Index’s data collection and computing infrastructure. The target is ambitious: 100x growth within the next year.
Index also functions as a labor marketplace where users and businesses can hire Creators for on-site tasks.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) ·