Memories.ai, a San Francisco startup founded by former Meta researchers, is launching Memories Upload, a gig-work platform that pays people to film their own hands at work so robots can learn from the footage. The company says it recorded more than 1 million hours of first-person video in the platform’s first month in beta and aims for more than 10 million hours by March 2027, a tenfold jump by our arithmetic.
The launch was first reported in a Fortune exclusive on 10 November. Memories.ai says its customers include “three of the top five frontier AI labs”, which it declined to name, citing non-disclosure agreements.
Business Pill · Physical AI Explained
An evergreen Business Engineer explainer on physical AI. It teaches the general idea only and does not discuss Memories.ai.
The key insight: Memories.ai says it recorded more than 1 million hours of first-person video in Memories Upload’s first month in beta and aims for more than 10 million by March 2027, ten times as much, by our arithmetic. It says three of the top five frontier AI labs are customers.
Contributors take paid video-recording tasks, Fortune reports, with cameras capturing “the hands-on details of preparing food, stocking shelves and repairing equipment”. Crypto Briefing reports payouts of $6 to $18 per task, with examples that include vegetable prep and bike repairs; Fortune does not give a rate.
The camera is the company’s own LUCI Pin, which “weighs under 45 grams and attaches magnetically to mounts worn on a personโs head, chest or wrist”. Two cameras can be paired for stereo capture, which helps estimate depth, according to Fortune.
Hours of first-person video Memories.ai says it recorded in its platform’s first month in beta and its goal for March 2027, both stated as ‘more than’, as reported by Fortune. The 10x multiple is our arithmetic.
What Labs Are Buying
Co-founder and CEO Shawn Shen told Fortune that customers increasingly want footage covering a wider range of work environments, tasks and camera angles. Developers also ask for wrist-level footage, to see the moment a hand touches an object.
The company labels the footage with Nvidia tools. Its video-captioning model is fine-tuned from Cosmos, Nvidia’s family of models for understanding the physical world, and the descriptions are tied to specific moments so developers can search for an action. Rick Zhuang, director of product and engineering for Cosmos at Nvidia, which works with Memories.ai, told Fortune: “What stands out is the focus on making everyday human experience searchable and useful for robot learning”.
Memories.ai has raised $16 million to date, Fortune reports, including an $8 million seed round in July 2025 led by Susa Ventures, with investors including Samsung Next, and an $8 million “seed extension” in June this year.
Camera wearers and coworkers who regularly appear must sign consent forms before recording, the company said. Before footage leaves the country where it was captured, automated tools obscure other people and mask details such as badges, screens and license plates, and staff manually review a sample of frames from each batch.
The Structural Read
As we read it, the business is a supply chain for physical-world data: Memories.ai developed its own camera, pays the contributors and uses Cosmos-based models to label the footage it supplies. The 10 million-hour goal is the company’s own target, and Fortune gives no independent count.
Rick Zhuang, director of product and engineering for Cosmos at Nvidia, to Fortune, 10 November 2026
“Recognizing an activity is different from understanding exactly how a hand grips a tool, moves an object or completes a step, especially when the camera is moving and objects are partly hidden”
Three Implications
FOR ROBOTICS DEVELOPERS Fortune reports that customers want footage across more work environments, tasks and camera angles, including wrist-level footage.
FOR CONTRIBUTORS Crypto Briefing reports payouts of $6 to $18 per task; consent forms are required for camera wearers and coworkers who regularly appear, the company said.
FOR THE DATA MARKET Fortune notes that Figure’s Index platform also pays people to record everyday tasks, and Physical Intelligence has tested human video with teleoperation data.
The framework reads: “Each moat is built by a different data pipeline.”
As we read it, first-person video of human work is one such pipeline, collected outside the public web. The framework predates this launch and does not discuss Memories.ai.
What Is Not Established
The 1 million hours and the 10 million-hour goal are the company’s own figures; Fortune does not give an independent count. The three frontier labs are unnamed. The $6 to $18 pay range comes from Crypto Briefing and we did not find it in Fortune’s report. No price for the data, and no revenue figure, was given in the reports we read.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
Memories.ai is paying gig workers to record first-person video of physical work and selling it, labelled, to robotics developers and, it says, three of the top five frontier labs. Its stated goal of more than 10 million hours by March 2027 is ten times the more than 1 million hours it says it recorded in its first month in beta, by our arithmetic.
A note on sourcing. Facts and quotations are from Fortune’s exclusive of 10 November 2026; the per-task payout range is from Crypto Briefing. Hour counts and customer claims are the company’s own. The tenfold multiple is our arithmetic. We did not contact Memories.ai.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
