This week has me thinking a lot about origin stories. Perhaps it’s because one of the robots featured below is nicknamed “Superman” (a name I’m guessing won’t make it too far beyond the demo stage).
Honestly though, that particular origin story is something along the lines of, “We spent three weeks building this robot that can run fast and jump high, and unrelatedly, our company is now worth $50 billion on the Shanghai Stock Exchange."
This week it’s the human origin stories that keep pulling me back in. Agility cofounder Damion Shelton tells us about becoming CTO of his first company on the walk home from one particularly bad job interview at Google.
Then there’s Samuel Reeves’s origin story, wherein he started FORT after DARPA bought a bunch of wireless e-stops from his robotic landmine-clearing startup. Oh, and give this week’s pod a listen to hear Nic Radford talk about being a high school high jump athlete prior to ending up at NASA.
Speaking of origin stories, don't forget to submit your job listings. I know you're hiring out there.
From 2012 to 2015, DARPA set the stage for robotics’ next decade. Teams from across the world built “human-supervised ground robots capable of executing complex tasks in dangerous, degraded, human-engineered environments.” Like the autonomous driving competitions before it, the three-year DARPA Robotics Challenge pushed the state of the art and kickstarted industries.
Another key precedent set by predecessors like the Grand Challenge was the importance of centering safety. The stakes may be lower, generally speaking, for a humanoid in an industrial setting versus an autonomous car on city streets, but both scenarios carry their share of bodily consequences.
When organizers wanted to standardize safety among the robots, they turned to Humanistic Robotics (HRI). The Pennsylvania startup initially crossed DARPA’s radar courtesy of U.S. Army funding for its efforts to automate landmine clearance.
“DARPA comes in and says, ‘hey, this wireless e-stop you built for yourself -- this seems useful for our teams,’ ” HRI cofounder Samuel Reeves explains. “They bought our wireless e-stops for every team. Every team was required to use it.”
It was a lightbulb moment for Reeves. The government agency had singled out one specific piece of the startup’s hardware and software portfolio. “That opened our eyes to ‘this experience we've gone through, where I built a robot, it's dangerous. What do I do about safety?’ was going to be a common experience that many other companies were going to have.”
Reeves sold his shares in HRI in exchange for underlying safety IP that would form the foundation of his next company. FORT was founded in 2018 to provide that key early step for other robotics companies, allowing them to outsource pieces of the necessary safety layer, so they could focus resources on other aspects of their tech stack.
“The reason I started my first company was that I interviewed with Google and the interview really didn't go well,” Damion Shelton explains. “When I was walking out of that interview, I got a call from a local entrepreneur who was coincidentally looking to commercialize my lab's work. He was like, ‘Why don't you just come be CTO?’ I said, ‘Okay, sounds like a good idea.’ ”
He adds dryly, “I didn't know what I was getting into.”
That turned out to be eight years as founding CTO of threeRivers 3D, Inc., a firm based just outside of Pittsburgh specializing in “practical and theoretical design of 3D scanner technologies.”
He was better prepared the second time around, jumping headfirst into his next startup, this time with fellow Carnegie Melon University (CMU) alum, Jonathan Hurst.
In 2015, Agility spun out of Hurst’s bipedal robot research at Oregon State University. He took the CTO title this time, as the startup established itself in Salem, Oregon. Shelton became CEO, maintaining his base of operations in Pittsburgh.
It’s a position he would keep for nine years before former Microsoft/Magic Leap executive Peggy Johnson stepped in as the company looked to commercialize and scale its efforts. Shelton spent the next year as Agility’s president, before stepping away from day-to-day operations last April (he remains board chairman).
He tells me he’s devoted much of the intervening time to government engagement and policy outreach, traveling back and forth to Washington D.C., as the U.S. hopes to build a cohesive robotics and AI strategy. Earlier this week, CMU announced Shelton’s appointment as associate VP and executive director of its Swartz Center for Entrepreneurship.
Founded in 2015 with a grant from Accel cofounder, Jim Swartz, the center describes its mission as, “work[ing] with its partners to serve the entire CMU community — departments, colleges, centers, and campuses — to accelerate bringing research innovations and promising ideas to the global marketplace and helping all entrepreneurial students, faculty, staff, and alumni [to] tap into the ‘innovation ecosystem.’ ”
Prior to cofounding humanoid startup Persona AI, CEO Nic Radford was a NASA roboticist who helped lead the development of Robonaut 2 and Valkyrie. Watch on YouTube>
Kathryn Zealand / Jessica Bath - Kathryn Zealand joins us to discuss her Google wearable robotics spinoff, Skip. After that, UCSF assistant professor, Jessica Bath discusses how exoskeletons are being used by Parkinson's patients.
Ali Agha- Live at Automate 2026, FieldAI's CEO discusses his NASA background and the state of physical AI.
Samantha Johnson (Tatum Robotics)- Johnson's ambitions to connect the deafblind grew as the community was cutoff from contact during the pandemic.
Unitree made its long-awaited, multi-billion-dollar market debut Wednesday. Shares in the Hangzhou-based humanoid maker surged up to 629%, before closing the day out at 460%. In USD terms, the stock shot from $22 to $163, before settling at just over $125 a share by the time the Shanghai market closed. At end of day, the company held a value of around $50 billion — down from $66 billion. As metaphors go, Unitree’s Monday demo was perhaps a touch on the nose, when it offered up a preview of “Superman,” a humanoid that can jump 6.5 feet from a standing position and hit speeds of more than 28 miles an hour. “This new machine has only been in development for a little over three months,” the company noted on social media, “with significant room for further improvement in the coming months.”
Amazon Wednesday announced the latest expansion of Prime Air to 500 U.S. cities and towns by year's end — 6x the drone delivery program's current scope. The Chicago, Syracuse, Cleveland, Atlanta, and Boise metro areas are listed as "coming soon." The drones can move products weighing under five pounds that take up less space than "a large shoebox," which the retail giant says covers around 60% of frequently purchased items. Most deliveries generally arrive in under an hour, per the company. Drone delivery is $5 for non-Prime members, $3 with Prime, and free for Prime members on purchases over $50.
Big week for people who like getting food dropped on them from flying robots. (Whomst among us?) Uber struck a deal to make deliveries via Zipline drones. The service will initially roll out in existing Zipline markets, Houston and Dallas, adding more as the drone delivery service enters additional cities. The companies are already tossing out big numbers. The linked press release says "Zipline and Uber plan to make one million deliveries a day, and have goals to expand far beyond that." Further down, Zipline notes that it has already delivered 20 million items, while flying its drones more than 135 million drone miles (basically regular miles, but for drones).
Serve signed a deal with Grubhub to offer robot food delivery in Chicago, Los Angeles, and Alexandria, Virginia (just south of D.C.). Serve-based Grubhub deliveries are also coming to San Jose, California. Miami, meanwhile, will see the launch of the first Serve microdepot, a “new and innovative class of small-footprint operating sites that handle robot staging, charging, dispatch, and maintenance without the build-out time of a full-scale facility.”
Meanwhile, Serve-owned Diligent Robotics has begun deploying Moxi 2.0 to various U.S. health care customers. The latest version of its bi-dexterous nurse-assisting robot features nine hours of run-time on a charge, an NVIDIA RTX A2000 processor, and "a new physical AI architecture." CEO Andrea L. Thomaz said: “Hospital hallways are some of the most dynamic environments a robot can operate in, full of fast-moving people, doors, and carts, all in tight spaces. To operate safely in that world, our first-generation platform often had to move conservatively. Moxi 2.0 is built with the compute, sensors, and models to reason about that complexity in real time, so Moxi can move with more confidence without compromising safety.”
Matic Robots gave me early access to a feature Mehul Nariyawala tells me the company has “been waiting for since 2017” (the year it was founded). The company's president drove the point home with a 30-second video from the fall of the following year with an early version running on a prototype named “Jupiter” that more closely resembled a coffeemaker on wheels. Existing customers get access to Cues today, via a firmware push. The feature is designed to untether the robot vacuum from the (very good) Matic app. Users can now direct the home robot to perform various tasks using voice commands in 75 languages, combined with gestures. The accompanying demo video features “follow me” and “return to dock” commands. I was also able to direct the robot to go clean a specific room, each Cue beginning with the “Hey Matic” wake word.
The robot’s display, coupled with some subtle body movements, alerts you that the system is listening, adjusting color and updating text to actions like “Following” and “Returning to Dock” when the action is being executed. Gestures add more granular Cues into the mix, such as “Hey Matic, clean this.” In a message ahead of the feature, Nariyawala notes, “Humans have understood each other through cues for 200,000 years. Academia has spent decades researching human-to-robot interaction, but we think the answer was always simpler: the best way to talk to a robot is the same way we talk to each other. A truly intelligent robot should learn how humans naturally communicate, instead of asking humans to learn yet another interface.”
Hot on the heels (hands have heels, too, it’s fine) of a $75 million funding round, Becca chats up Tacta Systems cofounder and CEO Vik Pavate. Hands are hard, we know this. Tacta is going the fluid-tendon route toward achieving high dexterity. Says Pavate, “The tactile hand, the sensors that went onto that hand, the congruences on the training system that allows the tactile data and vision data to be transferred there, the robotic foundation model and the special sauce that we have added on top of it in terms of encoders, in terms of reinforcement learning, the skill capture system, all of that specifically was designed from day one for high value manufacturing.”
This week, Liam’s got a piece about KAIST Hound, a quadrupedal research robot from Korea Advanced Institute of Science and Technology (KAIST) that utilizes AI to determine how to adapt to ever-changing terrain.
The Association for Advancing Automation (A3) is North America’s largest automation trade association representing more than 1,400 organizations involved in robotics, artificial intelligence, machine vision & imaging, motion control & motors, and related automation technologies.
.png?upscale=true&width=1120&upscale=true&name=Nic%20Radford%20(1).png)
-2.png?upscale=true&width=1120&upscale=true&name=Untitled%20design%20(5)-2.png)
.jpg?upscale=true&width=1120&upscale=true&name=matic%20cues%20800(1).jpg)
.png?upscale=true&width=1120&upscale=true&name=Tacta%20Hand%20Lift(1).png)
.jpeg?upscale=true&width=200&upscale=true&name=images%20(11).jpeg)
