Humanoid makers say the hands are the hard part. OpenTraining pays skilled workers to record how theirs do it, captured privately on smart glasses.
SHERIDAN, WY, UNITED STATES, October 20, 2026 /EINPresswire.com/ — On the eve of quarterly results from Tesla and Meta, the founder of physical AI data platform OpenTraining says the industry’s hardest problem is hiding in plain sight on every building site in the world.
In an interview this month, Elon Musk described building human-like robot hands as “shockingly difficult,” singling out tasks such as driving a self-tapping screw, which demands substantial force and fine control at the same time. Analysts have flagged hand dexterity as a key production bottleneck for humanoid robots.
“An electrician drives a self-tapping screw a few hundred times a day without thinking about it,” said George Bancs, founder of OpenTraining. “That is the skill every humanoid program is chasing. It does not live in a lab. It lives in the hands of tradespeople, and right now none of them are being paid for it. We pay them.”
Smart glasses as the training sensor for physical AI
OpenTraining equips skilled workers with smart glasses to record real work from a first-person view: the grip, the angle, the force, the order of operations, and the hazards an experienced worker sees coming. Each capture becomes structured training data for robot foundation models.
Workers are paid for every approved contribution and receive a share of downstream dataset licensing. Every capture is consented at the point of recording, reviewed before it enters a dataset, and stored with its consent record and provenance chain attached.
“Millions of people are now wearing cameras on their faces,” Bancs said. “That is the largest untapped robot training sensor network on the planet, but only if privacy is solved in the hardware rather than promised in a policy.”
Privacy built into the device, not promised afterwards
Smart glasses are under scrutiny on several continents. US federal and state courts have restricted them in court facilities, European regulators are examining the category, and Australia’s Privacy Commissioner opened a formal investigation this month as a bill to restrict imports was introduced in the Senate.
OpenTraining’s response is architectural. Face detection and pixelation run on the device, before anything is written to storage, so unredacted footage never leaves the glasses and is never seen by a human reviewer. License plates and documents go through the same pipeline.
“Anonymizing later is not privacy. It is a promise,” Bancs said. “We built it so that trusting us is not required.”
A data shortage, not a hardware shortage
Humanoid makers are moving from demonstrations to production lines, warehouses and job sites. The industry’s main data source today is teleoperation: rigs costing $50,000 to $150,000 each, operated in laboratories on tasks chosen by engineers. OpenTraining argues that approach cannot capture the judgment of a worker with thirty years on the tools.
Its platform records two tracks from every capture: the task itself, and the safety knowledge inside it. Hazard profiles have been authored across sixteen industry verticals, starting with Building and Infrastructure, and the first tasks targeted are the most dangerous: confined spaces, working at height, and high-risk trade work.
“The companies racing to put humanoids into factories and warehouses all need the same thing,” Bancs said. “Clean, consented, first-person data of experts doing real work, at scale. The workers who generate that data should be partners in the economy it creates, not casualties of it.”
OpenTraining’s Training Economy program was featured on Australian national news network 7NEWS this month. The company is a member of the NVIDIA Inception program and is now opening conversations with humanoid developers, robotics labs and trade organizations in the United States.
About Open Technology LLC
Open Technology LLC owns and operates OpenTraining.ai, a training data platform for physical AI, and is a member of the NVIDIA Inception program.
Press Contact
Agentic Press Holdings
connect@agenticpressholdings.com
Legal Disclaimer:
EIN Presswire provides this news content “as is” without warranty of any kind. We do not accept any responsibility or liability
for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this
article. If you have any complaints or copyright issues related to this article, kindly contact the author above.
Media gallery
