OPEN LINK ↗
// 86d 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
86d ago
2026-03-16
PUBLISHED
86d ago
2026-03-15
RELEVANCE
6/ 10
AUTHOR
esporx
