I sell my household chore videos to robots for AI training, earning $6.60 per hour, which is below the minimum wage.

Wired reporter Reece Rogers strapped an iPhone to his forehead, spending a week recording first-person perspective videos of himself doing household chores, uploading them to the data collection platform Luel to train humanoid robots.
(Background recap: The secrets behind AgiX Robotics’ prospectus: humanoid robots shipped #1 worldwide; commercialization relies on school research institutions)
(Additional background: Can a robot wash dishes and sweep the floor just by opening its mouth? Figure has launched the AI model “Helix,” targeting a household chores revolution)

$6.6, below the U.S. federal minimum wage of $7.25. This is the hourly wage Wired reporter Reece Rogers earned for spending an entire week strapping an iPhone to his forehead and recording first-person videos of himself doing chores at home.

The employer is a data collection platform called Luel. Every action he takes—slicing cucumbers, pouring drinks, and putting dishes into the dishwasher—will be packaged and sold to AI companies that are training humanoid robots.

Rogers is well aware that he might be training something that will eventually replace humans. In his report, he describes feeling embarrassed and also feeling bored and monotonous throughout the process.

Absurdly strict filming specifications

As Rogers describes it, Luel’s technical requirements are extremely strict: the recording device must be worn on the head, and handheld recording is not allowed; the angle must be a horizontal wide shot; the resolution must be at least 1080p. The most critical rule is this: both hands must be visible in the frame for more than 95% of the time.

His first 5-minute video was immediately rejected by the platform because his hands were not prominent enough in the camera shot. After that, he adjusted his filming method and started repeating the same action: tying shoelaces—20 times in total. Making salad, scrubbing dishes, cutting cucumbers—every action required his hands to remain occupying the main part of the camera view.

In the industry, this kind of video is called “first-person perspective data.” Put simply, it means mounting the camera at a person’s eye level, so the robot can observe the physical world from a “human perspective”: how hands approach objects, how fingers apply force, and how the gaze shifts to the next task after completing an action.

Text descriptions cannot convey this information; only continuous first-person video footage allows the robot to understand how an action takes place.

This also explains why the platform’s requirements are so precise: both hands must be visible 95% of the time, not for aesthetics, but because the robot model needs to extract a complete sequence of hand motions from the videos. Once the hands disappear, the value of the footage for training drops significantly.

Pricing for data

The hourly wage Luel offers is $6.6, and each video costs about $1 to $2. In the gig economy, this figure is on the low end, but it is not the only reference point.

Larger-scale platforms offer much better terms. Rogers mentions that companies like DoorDash pay around $20 per hour for similar videos—more than three times Luel’s rate. The gap between the two numbers reflects different buyers’ judgments about how urgent the data is: the closer the robot companies are to commercial mass production, the more they are willing to pay for integrated data.

Behind Luel’s low pricing lies another layer of structure. The data it collects is not for its own use; instead, it is packaged and resold as a middleman to robot and AI companies. In this chain, the worker who provides the videos receives the lowest compensation, while the licensing fees paid by the final buyers may be higher by several orders of magnitude. Rogers’s record doesn’t just show his own hourly wage figures—it reveals how labor is being priced in an emerging data market that is taking shape.

These kinds of jobs are currently spread across multiple platforms, with no unified pricing standard and no labor-protection framework to bring them under regulation. $6.6 is below the federal minimum wage, but it is not illegal because the money is defined as “video sales revenue,” not “wages.”

The irony of the whole situation is that some of the robots that might replace you in the future are ones you train yourself. And the fee you’re charging right now is even lower than the legal minimum wage…