A robot-training startup wants to clean homes for free, as long as it can record the job.
The company’s pitch is straightforward: it will send people wearing head-mounted cameras into homes to do cleaning work. The footage would be used as training data for robots, which need examples of ordinary tasks before they can attempt them. In other words, a human still does the cleaning, while the camera turns the visit into a data-collection session. The bargain is not cash for footage, but a household service in exchange for access to a private space.
That is the part worth watching. Robot companies need data from real homes because kitchens, bathrooms, clutter, lighting, and human habits do not behave like neat lab demos. This also flips the usual data-labor arrangement: instead of simply paying people to capture training footage, the offer asks homeowners to cover the cost with access, consent, and whatever the cameras happen to see. Home video is not the same as footage from a street or a warehouse aisle; it can capture possessions, routines, layouts, and all the small details people normally do not hand over to a startup.
The robots are still the future tense; the camera-wearing cleaners are very much present tense.