OpenTraining pays tradespeople to teach robots with smart glasses

OpenTraining says the humanoid robot bottleneck is not hardware but the lack of clean, consented first-person data from skilled workers. The company is now paying tradespeople to record real work on smart glasses and licensing that data to robot builders.

Why it matters: - Humanoid robot makers are moving from demos to factories, warehouses and job sites, but hand dexterity remains a core bottleneck. - OpenTraining is trying to turn skilled tradespeople into a paid data source for robot training, which could shift who gets compensated in the physical AI supply chain. - The company argues that better robot performance depends on capturing real-world judgment, force and sequencing, not just lab-controlled teleoperation.

What happened: - OpenTraining founder George Bancs said the company pays skilled workers to record how they perform real tasks using smart glasses. - The recordings capture a first-person view of grip, angle, force, order of operations and hazards. - OpenTraining says each approved capture becomes structured training data for robot foundation models. - Elon Musk said in an interview this month that building human-like robot hands is “shockingly difficult,” citing tasks such as driving a self-tapping screw. - Analysts have identified hand dexterity as a key production bottleneck for humanoid robots.

The details: - OpenTraining pays workers for every approved contribution and gives them a share of downstream dataset licensing. - Each capture is consented to at the point of recording, reviewed before entering a dataset, and stored with consent records and provenance information. - The company says face detection and pixelation run on the smart glasses before anything is stored. - OpenTraining says unredacted footage never leaves the glasses and is never seen by a human reviewer. - License plates and documents go through the same on-device processing pipeline. - The company said its platform records two tracks from each capture: the task itself and the safety knowledge inside it. - OpenTraining says hazard profiles have been authored across 16 industry verticals, starting with Building and Infrastructure. - The first tasks targeted are confined spaces, working at height and high-risk trade work. - OpenTraining says its approach is designed around hardware-based privacy rather than post-processing promises.

Between the lines: - The company is betting that the most valuable training data for humanoids lives with experienced tradespeople, not in robotics labs. - That framing also creates a labor-market argument: the people whose skills train robots should share in the upside. - The privacy pitch is central because wearable cameras face regulatory scrutiny in the U.S., Europe and Australia. - OpenTraining is positioning itself against teleoperation, which it says is expensive, lab-bound and too narrow to capture expert judgment.

What's next: - OpenTraining says it is opening conversations with humanoid developers, robotics labs and trade organizations in the United States. - The company’s Training Economy program was featured on Australian national news network 7NEWS this month. - OpenTraining is a member of the NVIDIA Inception program. - Open Technology LLC owns and operates OpenTraining.ai, the company’s physical AI training data platform.

The bottom line: - OpenTraining is trying to make skilled human labor the raw material for humanoid robot training, while making privacy and worker pay part of the product.