The robotics training data gold rush just got a new miner, and they're paying humans to wear sensors while they work.
The Summary
- Mecka AI raised $60 million in Series A funding led by crypto VC Framework Ventures to scale its human-data-driven robot training platform
- The company captures movement data from body sensors and iPhones, translating human actions into robot training datasets
- Mecka AI projects $100 million in annual run rate, signaling commercial traction in the robotics foundation model race
The Signal
Mecka AI is selling picks and shovels to the robot builders. While Tesla films factory workers and Figure AI burns compute on humanoid demos, Mecka took a different bet: the bottleneck isn't robot hardware or even AI models. It's training data that actually maps to real-world tasks.
The company instruments humans with wearable sensors and uses iPhone cameras to capture how people actually move through physical work. Not simulated. Not animated. Real humans doing real tasks, generating the ground truth data that foundation models need to generalize across different robot bodies and environments.
"The robotics industry is learning what computer vision learned a decade ago: you can't fake your way to generalization without massive, diverse, real-world datasets."
Framework Ventures leading this round is the tell. This is a crypto-native firm betting that robotics training data becomes a tokenized commodity market. The playbook writes itself:
- Workers get paid per task captured
- Data gets verified and labeled on-chain
- Robot builders buy training sets like API credits
- Quality contributors build reputation scores
The $100 million run rate projection suggests Mecka already has commercial customers buying data at scale. That's not vaporware. That's Tesla, Amazon, or the Chinese humanoid manufacturers paying for verified human movement data because their internal capture systems can't scale fast enough.
The Implication
If Mecka hits that run rate, we're watching the emergence of a new labor category: professional movement models. People who get paid not for completing tasks, but for teaching robots how to complete tasks. The gig economy graduates from driving Ubers to training the agents that will replace Uber drivers.
Watch for Mecka's data licensing terms. If they're building toward decentralized ownership where workers retain rights to their movement data, this becomes the blueprint for ethical AI training. If it's work-for-hire with perpetual licenses, it's just Mechanical Turk with better sensors.