A video went viral in India about a month ago appearing to show a vast number of garment workers wearing tiny, head-mounted cameras while they worked in a dreary-looking factory. A widespread hunch was the technology the video depicted was a system for what’s known as * egocentric data collection*—gathering first-person footage of people in action to train AI models, in order to replace the workers with robots. But it wasn’t totally clear if the video was real, let alone if the footage would or could be used to replace the workers.
That remains unconfirmed, but a different company called Human Archive has just raised $8.2 million from “Wing Venture Capital, NVP Capital, Y Combinator, and angels from OpenAI, Nvidia, Google, Mercor, AfterQuery, BAIR, SAIL, Brad Boa, and Meta,” according to TechCrunch, and it’s doing something remarkably similar to what’s rumored to be happening in the viral video. The investment confirms, at the very least, that Silicon Valley has some faith in technologies like this.
Human Archive doesn’t appear to mount cameras on the heads of factory workers specifically, but the rest, it appears, is about right.
In a YouTube video created by Human Archive’s leadership, one co-founder, Rushil Agarwal explains that the company attaches cameras to workers in “residential homes, restaurants, hotels, construction sites, logistics, and industrial environments worldwide.” TechCrunch reports that Human Archive is operating 1,000 of these pieces of hardware. TechCrunch points to gig economy platforms in India as the partner companies whose workers are wearing the headsets, although it notes that Human Archive hasn’t named any specific companies. Human Archive is headquartered, it says, in China and San Francisco.
The purpose of Human Archive’s headsets—along with additional wearable cameras—Agarwal says in the video, is to build two giant datasets: one in 3-D, drawn from cameras attached to what looks like a relatively nondescript black visor, and a second dataset focused on hand movements, drawn from 2-D video taken by wrist-mounted cameras.
Today, Human Archive is announcing our $8.2M seed round to model human embodied intelligence.
— Raj Patel (@babugi28) Despite decades of research, we still barely understand ourselves. Our goal is to learn how humans interact with the world, and over the past 6 months, our team’s made enormous progress…[May 26, 2026]
Another co-founder, Raj Patel, explains in the video that these are meant to be “the foundational datasets required to model human sensimotor intelligence at scale,” and that the company’s larger ambition is “capturing the structure of human interaction in the physical world.”
So is the purpose of these datasets to build world models that can be plugged into humanoid robots and eventually just replace human workers altogether? Agarwal seems to want to get ahead of any such speculation in the video. Human Archive, he says, doesn’t want to be limited by “the current action space of robots,” but wishes to expand more broadly into an understanding of “embodied intelligence.”
Embodied intelligence is a term seen often in soft robotics, referring to a sort of synergistic view of intelligence as a triangulation between the brain (or metaphorical “brain”), the body, and the environment.
Nonetheless, automating labor is an obvious application for these datasets, and Human Archive seems to understand that. In an X post Tuesday, Patel wrote, “The world is changing, and a new labor market is emerging.” He and his colleagues, he wrote, “believe our technology will become foundational infrastructure for automating manual labor, increasing global abundance, and advancing our understanding of human intelligence itself.”