The gig workers who are training humanoid robots at home

A new global gig economy is emerging where thousands of workers in countries like Nigeria, India, and Argentina are being paid to record videos of themselves performing mundane household tasks. Companies like Palo Alto-based Micro1 are hiring these contractors to mount iPhones on their foreheads and film carefully choreographed actions—folding laundry, washing dishes, making beds—to create training datasets for humanoid robots. With robotics firms including Tesla, Figure AI, and Agility Robotics investing over $6 billion in humanoid development in 2025, there's surging demand for real-world movement data that's difficult to simulate virtually. Micro1 CEO Ali Ansari estimates companies now spend more than $100 million annually purchasing this physical behavior data.

Workers are vetted by an AI agent named Zara and earn $15 per hour—significant income in regions with high unemployment. While this work boosts local economies, it raises ethical questions about privacy, consent, and the psychological impact of repetitive task recording. The approach represents a paradigm shift inspired by large language models: just as ChatGPT was trained on internet text, researchers believe humanoids can learn physical interaction through massive video datasets. Other companies like Scale AI, Encord, and even DoorDash are developing similar programs, while China uses state-run centers with VR headsets, creating a competitive global race to capture the physical data needed for the next robotics breakthrough.