Looks like Thanksgiving is on a Thursday yet again (how’s that for some finely tuned pattern recognition?), which means the Automated newsletter is off next week. The good news is we’ve got another jam-packed edition for you, with plenty of dinnertime conversational fodder for when you’re avoiding discussing politics at all costs.
Last week’s news cycle was all about CoRL – and we’ve got a bit of spillover down below. Before that, we kick things off by talking to Fort Robotics CEO Samuel Reeves about the company’s plan to go public, and what it means to scale safety.
We’ve got a couple of great contributor pieces. Matthew returns to discuss world models with experts from Agility, AMD, and the CTO of video generation firm, Runway. In a similar vein, Liam dives into recent research from U.C. Berkeley and Ken Goldberg around the use of agentic AI in robot training (remember the 100,000-year data gap?).
Hello Robot CEO Aaron Edsinger is on the pod this week, discussing his decades-long humanoid research and what age-tech assistance really looks like in practice. Speaking of the pod, we are so close to our 2026 YouTube subscriber goal. Please smash that subscribe button, if you haven’t already. And then go tell me about your favorite robot videos from the year for a fun wrap-up we have planned.
Oh, and while I’m piling on the requests, let me know who we absolutely need to have on the pod in 2027. I’m also working on my event schedule, so this is a good time to let me know if you’d like me to moderate or host your thing.
Happy American Thanksgiving, Black Friday, Small Business Saturday, Cyber Monday, Giving Tuesday, and Wacky Wednesday to those who celebrate. Lots more news below. See you soon.
Let’s call it “the Kiva problem” – the acquisition that permanently shifted the paradigm for warehouse robotics and scarred a generation of customers in the process. After purchasing it 2012, Amazon became Kiva’s only customer, leaving everyone who had previously relied on the AMR firm in the lurch.
The deal ultimately had a massive knock-on effect that reached far beyond Amazon’s walls, directly giving rise to now-dominant players like Locus. But the transaction left a lingering bitter taste in the mouths of many that has continued to shape the industry in other ways.
Industry vet Samuel Reeves cites the phenomenon as he runs down the motivating factors behind Fort Robotics’ recently announced plans to go public.
“We have nearly 700 customers now,” the CEO explains. “About a third of them are very large companies. As many of them have gotten larger they've asked this question: ‘We're making big commitments to you. How can you make big commitments to us as a smaller company?’ ”
Risk is an inherent part of doing business with a young company. While the precise numbers are hotly contested, common belief holds that between 70-90% of VC-funded startups fail. Such longshot odds compound for less established (at least in this world) categories like robotics/physical AI.
Fort has mitigated much of that concern with real customers and actual revenue. The self-described “Trust Layer for physical AI” counts some of the industry’s top names among its customers, including Google DeepMind, DoorDash, Zoox, and Agility. According to a recent filing, the company’s revenue grew “52.5% year-over-year to $12.5 million in the first nine months of 2026.” Continue reading >
Hello Robot's CEO explains why Stretch has one arm and wheels, how designing around ordinary home surfaces can reduce complexity, and why reliable daily assistance is the test that matters. Watch on YouTube.
Daniele Pucci (Generative Bionics):From flying humanoids to skin sensors, we walk through the Istituto Italiano di Tecnologia professor's fascinating work.
Boris Sofman (Bedrock Robotics): We discuss one of robotics' more fascinating journeys: from desktop companion, Cozmo, to heavy construction machinery.
Rodney Brooks (Robust.AI): Live from Silicon Valley, the robotics pioneer makes his triumphant return to help us mark one year of Automated.
Roughly this time last year, Automated visited the TRI offices and had a blast with Erin McColl discussing the year’s top robotics videos. We talked the good, the bad, and the questionable. We’re planning something similar for 2026 – with a bit of a twist (more soon). In the meantime, I need your help choosing the top videos – could be the most impressive physical AI demos, the coolest robots, or just a bit of whimsy. Submit your favorites at the link below.
Somehow missed this among last week’s deluge of CoRL papers (and, you know, other news): NVDIA debuted the latest version of GR00T. The world action model family is designed from cross-embodied deployment, with form factors including humanoids, dexterous hands, and grippers. GR00T 2 predicts video frames and robot actions in parallel, opening the system up to diverse training inputs, including human demonstration, along with UMI handheld or wearable devices. "On 50 tasks never seen on the target robot, only in human video or on other robots, GR00T 2 succeeds 40% of the time zero-shot," NVIDIA writes. "Five demonstrations per task, under three minutes of robot data, raise that to 95% after a 30-minute fine-tune on a single RTX 5090."
GlobalFoundries is set to pilot Boston Dynamics' Atlas in its semiconductor factories, as part of an early adopter program. While still very much in the early stages, the announcement marks the first time the electric humanoid is being put to work beyond automotive manufacturing – a fundamental base for the company, thanks in no small part to its parent org, Hyundai. GlobalFoundries currently deploys Spot robots for "autonomous enterprise asset management" at select locations, including a four-year stint in Burlington, Vermont. Today's news finds that end of the partnership extending to locations including Malta, New York; Dresden, Germany; and Singapore. The chip-maker's Atlas tests will begin early next year, in order to "evaluate potential humanoid use cases, assess customer readiness and integration needs, and begin early testing and training for specific task execution, complementing GlobalFoundries’ global program to upskill its employees in AI."
Following years of rapid expansion, California regulators are pumping the brakes on Waymo. The Department of Motor Vehicles has barred Alphabet's self-driving cars from state highways, citing “imminent hazard.” The cars, which have become a fixture across Silicon Valley, have been the source of multiple recent traffic incidents. Waymo will continue to operate vehicles in city streets while it works out fixes for recent issues.
Another huge physical AI round: Genesis AI just raised a $500M round that values the Parisian/Bay Area startup at $3.5B. Premji Invest, Khosla Ventures, and Eclipse co-led the Series A, which also featured Durable Capital Partners, T. Rowe Price Associates, Inc., Bpifrance, HSG, NVIDIA, AMD Ventures, Zoom, Samsung Next, and Ericsson. Per cofounder/CEO Zhou Xian, “Our mission is to remove the physical limits on human ambition. To accomplish this, we’re working to build the world's best general-purpose robot, one that works reliably in the real world, day after day. That requires the intelligence, hardware, and everything that goes into training and testing the robots to work as one system. By building it all ourselves, we can obsess over every detail and improve the entire robot faster in one closed loop. This funding will accelerate our efforts to deploy our robots in the real world at scale.” The funding will go toward hiring, scaling data, and deploying the company's Eno model.
Always good to see some funding for medtech. This week it’s Remedy Robotics, with the $62 Million Series. DCVC led the round. Remedy’s flagship product is the N1, a software/robotic solution for endovascular procedures that includes imaging, simulation, and operation. Per the company, "The funding will support several major milestones, including a planned pivotal clinical trial in partnership with the University of California and advancement toward FDA approval for neurovascular intervention. Remedy will also deploy prototype N1 Systems across its network of clinical partners, including through its strategic partnerships with SVIN Mission Thrombectomy and the Australian Stroke Alliance, while continuing to advance the system's industry-leading AI capabilities and autonomous functions."
You may have missed it (it is really small, after all), but for the past decade or so, MIT has been doing a good bit of work in the microrobotics (how did we not go with “microbotics?”) space. This 244-mg robot stings like a butterfly, floats like a bee, and is now capable of landing on the surface of a sunflower outside of a laboratory setting. The system currently requires a tethered powered source, but the team is hoping to build future versions that can do, you know, bee stuff, like pollinating plants.
Thoroughly enjoyed this month's feature from contributor Matthew Smith, which dives into the complex and oft-misunderstood land of world models. Why, say, is it reasonably easy to get a robot to bust a dance move but not carry around a cup of water without spilling? Matthew spoke to Agility's Jonathan Hurst, Runway's Anastasis Germanidis, and AMD's Head of Robotics, Industrial, Healthcare, and Embedded Computing, Amey Deosthali, the latter of whom breaks down the distinction between simulation and world models thusly, "A simulator normally starts with an engineered representation of physics, geometry, sensors and objects. A learned world model starts from data."
This week, Liam breaks down some work coming out of Ken Goldberg's lab at U.C. Berkeley. If you've followed Ken in recent years, you're no doubt familiar with his "100,000-year data gap" thesis. A lot of venture capital has been pumped into paving that gap over. Discussing how agentic AI could be used to better train robotics, the professor noted that recent work gave him "goosebumps." This piece breaks down how the technology can train systems to perform a wide range of tasks ahead of deployment, from sorting packages to making sandwiches.
The Association for Advancing Automation (A3) is North America’s largest automation trade association representing more than 1,500 organizations involved in robotics, artificial intelligence, machine vision & imaging, motion control & motors, and related automation technologies.