REDDIT · REDDIT// 27d agoNEWS

Gig workers film home chores to train humanoid robots

A growing gig economy has emerged around recording household tasks — cooking, cleaning, folding laundry — to generate training data for humanoid robots. Companies like Sunain, Instawork, Scale AI, and Micro1 collectively employ tens of thousands of workers globally, paying $80–$1,200 per session to capture the physical AI data that Tesla, Google, and Figure AI need to train robots that can operate in real homes.

// ANALYSIS

Robotics hit the same data wall that LLMs hit years ago — and the answer is the same: pay humans to generate it at scale.

–Unlike text/image models, physical AI can't scrape the internet; every hour of real-world manipulation data must be physically performed and recorded
–Scale AI alone has amassed 100,000 hours of robotics footage; Sunain has 25,000 contributors globally — this is now industrial-scale infrastructure
–Goldman Sachs projects the humanoid robot market at $38B by 2035, making this data layer a critical and undervalued part of the stack
–The irony is sharp: workers in hospitality and domestic labor are funding the training of robots designed to replace them in exactly those roles
–China is running 40+ state-owned VR teleoperation centers for the same purpose — a direct signal this is a strategic infrastructure race, not just a niche startup trend

// TAGS

roboticsllmdata-toolsinfrastructureresearch

DISCOVERED

27d ago

2026-03-16

PUBLISHED

27d ago

2026-03-15

RELEVANCE

6/ 10

AUTHOR

esporx