Welcome, sci-fi realists…
Researchers at MIT tested a new imaging system that relies on sonar and visual data to map out dark and murky places underwater.
This “Sonar-MASt3R” system was attached to a robotic arm and tested in cloudy water to see if it could handle such environments (it can). It’s unclear what the full list of goals are for this endeavor, but the idea of having robots remove explosives from the dark areas ships pass through was thrown out there during the study. You, uh, might be able to conjure a possible use case for this, if you follow current events at all.
That all means this device could turn robots into powerhouses that save ships, study fish and possibly hunt for lost treasure. Sonar mapping isn’t new, but there’s untapped potential for underwater robots in rough areas. For now though, we can just imagine a robot strapped to a Sonar-MASt3R looking for valuable objects at the bottom of a lake.
Anyway, here's what you need to know about robotics right now:
Teradyne Robotics will display production-ready physical AI products this week.
Jeff Bezos’ new company, Prometheus, intends to build AI that designs robots.
LG Electronics is building a factory of training data in South Korea for humanoids.
The U.S. Congress looks to create a future National Commission on Robotics.
🐱Teradyne Robotics wants factories to meet physical AI before humanoids do
Teradyne Robotics is set to demo production-ready physical AI applications at Automate 2026 in Chicago this week. Translation: these are robots that are ready for industry use. They use AI to see, adjust, and act in the physical world instead of following every movement from a fixed script.
That is the unglamorous version of the current robotics boom, which also makes it the version factories can actually buy (and rely on). Teradyne owns Universal Robots, which makes collaborative robot arms, and Mobile Industrial Robots, which makes autonomous mobile robots that move material around facilities.
At Automate, Teradyne is showing a booth full of systems built for the messy parts of work:
MiR1200 Pallet Jack, a physical AI pallet jack that Teradyne says manufacturers can purchase now.
UR AI Trainer, built with Scale AI, which lets workers physically guide a robot through a task so the motion can train future robot models.
Cambrian’s cable demo, where dual robot arms use AI vision to identify and insert copper cables into dense server racks.
AICA’s polishing demo, where a robot learns a force-sensitive buffing motion from a single human demonstration.
Vention’s bin-picking demo, which claims a 99% first-pick success rate with a UR12e robot and 3D vision.
The event runs from June 22-25 in case you want to witness them in action yourself.
What this likely means for the future: Teradyne is competing against the humanoid hype cycle by selling boring, buyable automation that already fits factories. Factories need robots that notice a box has shifted, grab the right part anyway, and keep the line moving. Teradyne is packaging physical AI into tools manufacturers already understand: robot arms, pallet jacks, mobile carts, vision systems, and coordination software.
Meanwhile, humanoid companies are starting to rack up real pilots across two fronts: factories and the home. From Agility’s Digit in warehouses to Figure at BMW and Apptronik with Mercedes and Jabil. But most of that market still lives somewhere between “promising test” and “please watch this robot sort packages for 18 hours."
On the home-front, 1X already has its NEO robot in people’s houses (tele-operated for now), Sunday’s Memo is learning to clear tables, load dishwashers, and make espresso, and Physical Intelligence is building the robot “brain” layer that could make these machines less brittle.
Teradyne’s pitch is less theatrical and more procurement-friendly: robot arms, pallet jacks, mobile carts, vision systems, and coordination software that manufacturers already understand. If the demos land, buyers get a more concrete choice: automate a specific task now, then add AI adaptation where the real world gets messy. The humanoid waitlist will continue to wait for tasks that truly need a human-shaped machine.
Our FYI read: flashy demos are fun and help attract funding to the industry, but deployable robots are what turn robotics companies into line items in factory budgets (and actually keep the lights on for the bots AND their makers).
🕹️Robots in Action
A brief CNBC video below, in which some of Teradyne’s robots were showcased overseas in India fairly recently, demonstrates how it looks when they take on warehouse or factory operations.
As evidenced by the video, Teradyne’s product family seems to have no problem with:
Clawing cylindrical, rectangular, or square-shaped objects and placing them in slots.
Sticking suction cups on top of boxes and lifting those boxes for pick and place operations.
Transferring large crates (in the case of autonomous mobile robots or AMRs).
📰Around the Horn
Genesis AI just unveiled Eno (featured above), a wheeled general-purpose robot that aims for human-level capability without the full humanoid body.
AI robotics startup Theker raised $85M to build an industrial robot that’s reconfigurable for any factory task.
An American robotic drone performs the first ever robotic-led rescue at sea.
LG Electronics is building a new “training data factory” in Seoul, South Korea to train humanoid robots.
Jeff Bezos’ new company, Prometheus, will build AI to expedite the process of designing robots among other physical systems.
The U.S. Congress wants to establish a National Commission on Robotics through bipartisan legislation.
🚨Weekly Feature: Bots Behaving Badly
Now, what isn’t working so well right now?
As humanoid robots copy human activity and become more dextrous, it’s important to remember they aren’t made of human flesh and bone and don’t know their own strength. Case in point: humanoids are capable of playing soccer and kicking soccer balls so hard that they dent holes in walls or obliterate cameras used to record them.
The test subject Beijing’s Booster Robotics used didn’t intentionally dent the wall, but it couldn’t foresee its ability to do so. Not only that, it can’t feel remorse for the damage.
Amid the 2026 World Cup, researchers are floating the idea of a RoboCup, which now seems like a scary idea given this nugget of news. Imagine the balls flying off the field so hard and so fast they might dent your face. It’s difficult to imagine what more of this would look like on a soccer field without some changes. It might work in a more controlled environment with thicker walls (or, y’know, lock the bots in a giant steel mesh dome??), but who knows?
📡Industry Signals
Companies and researchers want to harvest training data so bots can learn from human examples. Clearly, they believe this is the quickest way to mass-produce humanoids.
Despite the hype around what humanoids are physically capable of (e.g. dance competitions), the real practical value for industries lies in the dexterity and range of motion of new humanoids.
Startups focused on both physical AI and robot products are securing millions, proving investors don’t see robotics is just hype anymore.
Hi! I’m Cameron Hashemi-Pour a technical writer and editor who covers AI, robotics, machine learning, Industry 4.0, cybersecurity, and emerging technology. Previously, I wrote for TechTarget/Informa TechTarget and contributed scripts for the “Eye on Tech” YouTube channel.
That’s all for this week!