Figure launches Index, a gig platform to train household robots

Humanoid-robot maker Figure AI launched Index, a consumer app that pays members of the public to film themselves doing everyday household and workplace tasks, feeding the resulting video into training for the company’s Helix robot-control model. Figure said its Creators — its term for paid contributors — had uploaded more than 16 million videos and been paid a combined $15 million, and that the company planned to spend more than $1 billion on data collection and compute over the following year. The app had been downloaded more than 264,000 times across over 100 countries, with more than 44,000 weekly active users.

Figure described a five-stage pipeline turning submitted video into training data — filtering, fraud review, deduplication, rebalancing and annotation — and said the platform was processing roughly 30 minutes of video uploads every second at peak. Beyond simply filming tasks, the app also supports on-demand jobs in which Creators complete a household or workplace task directly for a paying customer, with the recording itself as the deliverable.

Figure framed the effort as addressing a data bottleneck rather than a modelling one: real-world manipulation tasks vary too much between individual homes, objects and environments for simulation alone to cover, and the company said generalisation was “a data problem” more than an algorithmic one. The approach makes explicit a pattern already visible across humanoid robotics: because training data has to come from people actually performing physical tasks, progress increasingly depends on paid human labour rather than on compute or model architecture alone.

The launch extended what Figure, announcing a Series C round that valued it at $39 billion the previous September, had already described as a new data-collection effort using human video — at a point when investor and industry interest in embodied AI continued to outpace Figure’s commercial deployment of robots at meaningful scale.