A startup is offering free home cleaning, provided every room is recorded on camera.
The company sends human workers wearing head cameras into customers' homes to perform standard cleaning tasks. The workers are generating training data, not just tidying up. Their footage captures a human navigating real domestic spaces, picking up objects, and handling the variety of surfaces every home has. Homeowners get a clean house; the company gets a labeled demonstration dataset for teaching robots to do the same work.
Household robotics has a data problem that simulation cannot fully solve. Living rooms are not factory floors; every layout, floor type, and countertop arrangement is different, and models need exposure to real homes at scale to generalize. Trading cleaning labor for camera access costs less than operating dedicated research spaces or paying data collectors directly.
The "free service in exchange for your data" model is one of tech's oldest arrangements. It is just usually your browsing history, not a video survey of your floor plan and everything on it.