Free cleaning service for you: the condition is that the cleaner wears a camera into your home, MicroAGI feeds the next-generation household robot

German startup MicroAGI launches Shift app in New York offering free home cleaning in exchange for first-person videos shot by camera-wearing cleaners, used to train the next generation of household robots.
(Background: AI stock guru Serenity highlights four photonic stocks: AAOI surges 11x with expansion, XFAB soars 76%)
(Additional context: Arthur Hayes again calls for HYPE to hit $150! Declares TradFi as garbage)

Table of Contents

Toggle

• Two hours of cleaning, in exchange for home video data
• Simulated training that doesn’t teach grasping
• The promise of anonymization, and a question with no answer

Would you be willing to have your home videos exchanged for free cleaning? This week, German startup MicroAGI announced via its Shift app that it will offer free home cleaning to residents in New York. Cleaners wear cameras or smart glasses while working, and the first-person videos they record will be used to train next-generation household robots.

Today, we're launching shift. We're starting by cleaning your apartment in New York City, for free.

Here's how it works. Book a shift cleaning. A vetted shift operator comes to your home wearing one of our devices. They clean. They leave. You pay nothing.

In exchange, we record… pic.twitter.com/oBrCXcEz5G

— shift (@joinshiftX) May 28, 2026

Two hours of cleaning, in exchange for home video data

MicroAGI positions itself on its website as a “team of engineers, researchers, and operators accelerating embodied intelligence development.” Embodied AI, simply put, is the ability for machines to move and manipulate objects in the real physical world — a core bottleneck in current humanoid robot development.

The process of the Shift app is quite simple: users input their phone number, email, home address, and access info on the app, then schedule about two hours of free cleaning service. The cleaner wears a camera or smart glasses to record first-person perspective work videos. After uploading, the data is used to train robot models.

MicroAGI claims that in Q1 of fiscal year 2026, it paid over 10,000 “operators” more than $5 million across 15 countries. Currently, the service is limited to New York, but plans are underway to expand to San Francisco, London, Zurich, and Munich, with future scope extending from cleaning to plumbing, electrical repairs, and daily chores.

Simulated training that doesn’t teach grasping

Real-world first-person videos are currently irreplaceable training data, and the entire industry is trying to solve this “data drought” problem.

Scale AI has collected about 100k hours of robot training videos; DoorDash launched the Tasks app in March 2026, allowing 8 million U.S. delivery drivers to earn money by filming tasks like folding clothes, washing dishes, and making beds — deliberately avoiding states with stricter data privacy laws; Nigeria and India gig workers strap iPhones to their foreheads to record household chores, earning about $15 per hour, with at least 10 hours of footage required weekly.

The end goal of these data efforts is to build training datasets for humanoid robots from Tesla, Figure AI, Agility Robotics, and others.

The promise of anonymization, and a question with no answer

Shift’s FAQ states that all names, faces, and personal information are automatically anonymized before use; privacy policies further specify that the company performs “advanced machine learning models” directly on smart glasses or camera devices, executing “irreversible transformations” before uploading videos to the cloud, including automatic face blurring and recognition info masking. Screens, IDs, papers, and phone screens are all processed.

However, a key question remains unaddressed: Can users request their home cleaning videos to be deleted from the training dataset? The policy offers no clarification.

A more fundamental issue: anonymizing faces does not mean homes cannot be identified. Photos of interiors, files on tables, specific room layouts — all could contain clues for reverse identification. Embodied intelligence training relies precisely on these details: object positions, environmental structures, spatial arrangements. And these are the parts of a home that are hardest to truly anonymize.