An AI lab specialising in real-world data collection has found a new way to feed the hunger for robot training data: pay for people's home repairs. HUMXN is offering free plumbing, electrical, HVAC, air-duct cleaning and carpet cleaning to residents of three US cities, in exchange for letting a technician wearing "wearable capture technology" record how the work is done, according to a report by Tom's Hardware published on Friday.
The program launches in Minneapolis, Chicago and Miami — cities chosen, the company says, because their very different climates produce very different home-service conditions, from frozen pipes to humid-weather HVAC strain. Homeowners can check whether they qualify on the HUMXN Free Home Service webpage.
Why Robots Need to Watch a Plumber
The premise is simple and increasingly urgent for the robotics industry. Today's large language models were trained on text and images scraped from the internet, but there is no equivalent trove of data showing how skilled people physically interact with the world. Lab demonstrations, HUMXN says, struggle to capture the natural movements of experienced technicians working in cramped, unpredictable real environments.
"The next generation of AI is going to need more than text and images from the internet," HUMXN founder Thomas Unise said, according to Tom's Hardware. "Robots need to understand how skilled people interact with the physical world. Home-service technicians perform millions of highly technical tasks every day. We are creating a way for homeowners and service companies to benefit financially from the value of that expertise."
For residents, the trade is straightforward: a free repair that would normally cost hundreds of dollars, in exchange for a recorded visit. For technicians and service companies, HUMXN frames the program as a revenue source — a way to be paid for expertise that AI labs previously had few legitimate ways to obtain.
To follow the latest AI developments in robotics, models and the data pipelines behind them, AI Buzz Wire covers the full pipeline from collection to deployment.
Paying for Data in an Era of Scraping Scandals
The timing is notable. Major AI companies — including Meta, Nvidia and OpenAI — have faced repeated accusations of using copyrighted or personal data without permission to train their models, and courts on both sides of the Atlantic are still sorting out where fair use ends. Against that backdrop, a lab that literally subsidises home repairs to obtain consented recordings is betting that licensed, real-world data is worth the logistics.
The approach also reflects a broader shift in what AI companies consider valuable. As text-hungry language models saturate the available internet, the frontier is moving toward embodied AI — robots that can cook, clean, repair and navigate homes. That market needs video of humans doing exactly those tasks, in real homes rather than staged demos, and wearables that passively record a technician's hands, tools and point of view are currently the cheapest way to get it.
The five services on offer are not random, either. Plumbing, electrical work and HVAC repair consistently rank among the most physically complex tasks a mobile robot could attempt: they require fine manipulation in tight, poorly lit spaces, constant improvisation around unfamiliar layouts, and judgment calls that are easy for an experienced human and brutal for a machine. Carpet cleaning and air-duct work add repetitive, large-area motion patterns that map more cleanly onto what today's robots can already approximate. Capturing both ends of that difficulty spectrum in a single dataset is precisely the point.
Open Questions: Consent, Privacy and Scale
The program raises questions the industry has not fully answered. Homeowners consent to the recording, but the footage captures technicians — and homes — in intimate detail, and the company's policies on how long recordings are kept, who can access them and how they are anonymised will shape whether the model is seen as fair compensation or a new form of surveillance dressed up as a discount.
There is also the labour question. Skilled trades are among the jobs analysts have long considered safest from automation precisely because they demand dexterity in unpredictable spaces. If labs like HUMXN succeed in capturing enough of that expertise on video, today's well-paid repair work becomes tomorrow's training corpus — with technicians, in effect, helping to build the systems that may one day compete with them.
For now, HUMXN's offer is live in three cities, and the company's wager is that free carpet cleaning is a small price for the data that teaches a robot to do it forever. Whether homeowners agree — and whether technicians do — will be the first real test of the pay-for-physical-data model at household scale.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
