This startup will clean your apartment for free, if AI can watch

Shift Robotics, a consumer-facing brand of microagi, is providing free apartment cleaning services in New York City as part of a data collection effort to train AI systems for physical tasks. Cleaners wear head-mounted cameras to record footage of chores like dishes, mopping, and laundry, which is anonymized and sold to AI labs and robotics companies. The startup’s launch video went viral with over 8 million views, and its initial 250 cleaning sessions sold out almost immediately. Microagi, founded in 2023 by former Formula One engineers Bercan Kilic and Yoan Iliev and AI researcher Anton Poletaev, operates as a research lab based in Germany. The company aims to develop 'end-to-end physical AGI,' enabling machines to perform real-world tasks autonomously. Shift currently operates in 15 countries with 14,000 operators collecting data to accelerate AI training. The free cleaning offer addresses a critical challenge in robotics: the lack of real-world training data. Unlike large language models, which rely on text and images from the internet, robots require physical examples of tasks like cleaning or cooking. Shift’s model pays gig workers to record these activities, with faces and screens blurred for anonymity. Early users reportedly spread the word by offering cleaning services to neighbors, expanding the data collection effort organically. Shift’s US General Manager, Harry Kilberg, stated that the company’s unit economics are strong due to high-quality data processing, allowing it to command premium prices from AI labs. The startup plans to expand across the US, adding free or subsidized services like cooking and plumbing. Kilberg emphasized that while household robots remain years away, the data collected now will be essential for their future development. The initiative reflects a broader industry shift toward real-world data collection, with companies like Scale AI and micro1 transitioning from chatbot training to physical AI applications. Shift’s approach leverages human labor to bridge the gap until robots can perform tasks independently.