Hands That
Teach Machines
A decentralized network capturing real-world human demonstration data for robot intelligence.
Every task performed. Every motion recorded. Every contribution rewarded.
How It Works
From human motion to machine capability
Capture
Distributed operators perform everyday manipulation tasks wearing lightweight capture hardware. Kitchens, warehouses, hotels, homes: real environments with real physics, real clutter, real variation. No robots required at collection time.
Process
Every session is automatically synchronized, quality-scored, and packaged into training-ready formats. The protocol filters noise, detects occlusions, segments tasks, and ensures only high-fidelity data enters the network.
Train
Frontier robotics labs and foundation model teams access the most diverse real-world manipulation dataset available, spanning environments, tasks, and human variation at a scale no single lab can build alone.
Join the Network
Two paths, one network
Your work trains
the future
If you work with your hands, you already have what the robotics industry needs most. Join the network, capture data during your normal tasks, and earn for every quality session you contribute. No robotics experience needed.
Apply to ContributeReal-world data,
training-ready
Synchronized RGB, depth, and inertial data captured across hundreds of real environments by distributed human operators. Quality-scored, annotated, and delivered in standard formats.