The Sashimi robot is real and it doesn’t fumble at slicing and dicing
A Norwegian research team built a robot that can slice and serve salmon sashimi using three arms, AI training, and a tactile sensor that knows when the blade hits the board.
Kass
Three arms, a GelSight sensor, and surprisingly clean chopstick work.
npj Robotics / Nature
Robots can pick up boxes, sort packages, and screw in bolts without breaking a sweat. Some of them can even walk and run like humans. Hand one a floppy, slippery piece of raw salmon, though, and everything starts falling apart.
A team at the Norwegian University of Science and Technology set out to solve that problem. The result is the Sashimi-Bot, a three-armed robot that can prepare sashimi from a raw salmon loin without a chef in sight.
npj Robotics / Nature
So how does the Sashimi-Bot actually work?
It divides the job neatly between its three arms. The first arm stabilizes and positions the salmon on the cutting board. The second holds a chef’s knife and slices. The third picks up each finished slice with chopsticks and transfers it to a serving tray.
What makes this more than just clever arm arrangement is how the robot learned to do it. Lead researcher Sverre Herland and the team trained it using deep reinforcement learning inside a virtual simulation.
The technology let the robot practice thousands of movements and learn through trial and error, without any practice on real fish.
npj Robotics / Nature
Does it actually hold up?
The knife arm also carries a GelSight tactile sensor, a soft gel surface with an embedded camera that tells the robot exactly when it has reached the cutting board.
During testing, the robot cut 34 slices of salmon. It successfully grasped 26 of the 28 slices that fell onto the cutting board with chopsticks. An additional six slices that had stuck to the knife blade were retrieved directly from it.
Each cut cycle averaged 27.9 seconds. The study is published in npj Robotics (via TechXplore). While most robots do best with rigid, predictable objects, the Sashimi-Bot is more significant than its culinary application suggests.
It is an example of robots handling delicate, irregular materials by making real-time movements and adjustments.
For more than five years, Shikhar has consistently simplified developments in the field of consumer tech and presented them…
macOS 27 means the end of the Hackintosh Era, but does anyone still need one?
Apple's latest software update shuts the door on x86, but today's Macs have already made the original Hackintosh dream surprisingly accessible.
While the tech world was busy obsessing over Liquid Glass, smarter Apple Intelligence features, and all the shiny new additions arriving with macOS 27 Golden Gate, Apple quietly slipped in another announcement at WWDC 2026 that didn't get nearly as much attention. Buried in the compatibility list was a simple but significant detail: Intel Macs are no longer supported. For millions of users, that's just another software update requirement. For a passionate corner of the internet that has spent nearly two decades bending technology to its will, it's something far bigger. It's the end of the traditional Hackintosh era.
If the term sounds unfamiliar, here's the quick version. A Hackintosh is a regular PC that's been modified to run macOS instead of Windows or Linux. Using community-developed bootloaders such as OpenCore and carefully selected hardware, enthusiasts managed to convince Apple's operating system that it was running on a genuine Mac. The process was anything but straightforward, but for many, that challenge became part of the fun.
Instacart is testing camera-ready AI shopping carts that sound convenient, but equally scary
Caper Carts promise faster checkout and personalized savings, but the cameras, location tracking, and on-cart ads make grocery shopping feel a lot less private.
Instacart's AI shopping carts are moving into select Weis Markets stores in Pennsylvania, with more locations planned this year. On the surface, the Caper Cart upgrade sounds useful, since shoppers can see a running total, clip digital coupons, use loyalty rewards, weigh items, and pay from the cart.
The privacy tension comes from the hardware needed to make that work. The carts include basket-facing cameras, outward-facing cameras, location-tracking systems, scales, touchscreens, and payment terminals, turning an ordinary grocery basket into a rolling sensor platform.
OpenAI wants an all-knowing personal AI agent for everyone on Earth
The company is framing personal AGI as the mass-market endpoint of its AI race.
OpenAI is laying out a future where advanced AI reaches billions of people, not only the companies and governments racing to control it. Its latest plan centers on an AI for everyone, a personal AGI that would work as a deeply capable assistant for daily life, work, and discovery.
The company calls this its third phase. After proving the technology could work and turning it into products used at scale, OpenAI now wants to make powerful AI broadly available while pushing systems that can accelerate science and economic growth.
