Figure AI has unveiled Index, a crowdsourced platform for building what the company calls the largest useful robot training dataset in the world. The platform, which operated in stealth for four months before today's announcement, has already accumulated 16 million video uploads from users across 108 countries.
Introducing Index
— Figure (@Figure_robot) August 25, 2026
Today we're coming out of stealth with the largest and most diverse robot dataset to scale general purpose robots
To date we've crossed 264,000 app downloads and uploaded 16 million videos pic.twitter.com/1J1CCgbW6W
According to the company, Index crossed 264,000 app downloads during its stealth period, with over 43,000 weekly active users now contributing data. Figure says it has paid out $15 million to contributors so far, and the platform is uploading more than 30 minutes of video every second.
The premise is straightforward: Figure AI, founded in 2022 by Brett Adcock, builds humanoid robots that need to learn how to operate in the physical world. Unlike large language models, which can train on internet text, embodied AI systems require real-world demonstrations of tasks like opening drawers, folding laundry, and navigating cluttered spaces. That data simply does not exist on the internet at scale.
The Data Problem in Robotics
This has been the central bottleneck for the entire humanoid robotics industry. Language models scaled on internet data. Embodied AI must build its data from the physical world, and that changes everything. Traditional approaches rely on teleoperation by expert operators or simulation environments, both of which have significant limitations in scale and real-world applicability.
Figure has already been exploring data collection at scale through its Project Go-Big initiative. The company is building what it describes as the world's largest and most diverse humanoid pretraining dataset, accelerated by an unprecedented partnership with Brookfield, which owns over 100,000 residential units worldwide.
But Index represents a different strategy: instead of capturing data through corporate real estate partnerships, Figure is paying ordinary people to record themselves performing everyday tasks. The company has committed to spending over $1 billion on data and compute over the next 12 months.
Why This Matters for Labor Markets
Figure is not shy about its endgame. The announcement explicitly states that Index is laying the groundwork for ordering robots as a service, envisioning a future where robots handle household tasks currently performed by human workers. This is a significant development for the labor market, but not necessarily the dystopian one that critics often predict.
By 2026, Figure's robots have demonstrated the potential ability to perform household work. The Figure 02 robot supported the production of more than 30,000 BMW X3 vehicles over 11 months in South Carolina. CEO Brett Adcock has sketched out a home lease model of roughly $400 to $600 a month.
The optimistic case is that robots handling repetitive physical tasks could free human workers for higher-value activities, address labor shortages in sectors like elder care and manufacturing, and make services more accessible to people who cannot currently afford them. A robot that costs $500 per month to lease could provide household assistance to millions of families who cannot afford human help.
Edge Computing and Decentralized Intelligence
What makes Index particularly interesting is its distributed architecture. Figure is essentially building a global network of edge contributors who capture training data in their own environments. This addresses one of the fundamental challenges in robotics: the long tail of environmental variation. A robot trained only on data from Silicon Valley offices will struggle in a Tokyo apartment or a rural farmhouse.
By collecting data across 108 countries, Figure is building robustness into its models from the start. The diversity of contributors creates a dataset that captures cultural, architectural, and material variations that would be impossible to simulate or capture through centralized collection.
Figure's Helix system learns from three data sources: human demonstrations captured through teleoperation, real-world robot performance logs, and internet-scale video data of humans performing everyday tasks. Index appears designed to massively scale that third category.
The Competitive Landscape
In late 2025, Figure AI raised more than $1 billion in Series C funding at a valuation approaching $39 billion, nearly 15x as much as the year prior's valuation. The company's competitors in the humanoid space are pursuing their own data strategies, but none have announced anything at the scale of Index.
Figure launched the Figure 03 robot in October 2025, and is now doubling monthly shipments with the BotQ factory producing one robot every 90 minutes as of April 2026. The company reached 1,000 Figure 03 units manufactured on July 23.
The timing of the Index announcement suggests Figure is entering a new phase. The company has proven it can build robots that work in controlled industrial settings. Now it needs the data to make those robots useful in the chaotic, unstructured environments of everyday life.
What Happens Next
Figure's path to 100x, as the company frames it, depends on whether crowdsourced data collection can actually produce the quality and coverage needed for general-purpose robots. The $15 million already paid out to contributors suggests the economics are working at current scale. Whether those economics hold at 10x or 100x remains to be seen.
The implications for how humans and machines share work are significant. If Index succeeds, Figure will have built not just a dataset but a distributed labor market where people are paid to teach robots how to do their jobs. That is a strange loop, and one worth watching closely.
For now, the platform is live and accepting contributors. Figure has published a detailed writeup on its website, and the global contribution map updates in real time.
