In the original visual context, the robot remembers the past.
SimpleMemVLA: Context in native visuals acts as a memory, allowing robots to remember the past. Without adding dedicated memory modules, and directly using the time-lapse...
Global Robotics News
A focused editorial channel for robotics companies, products and technologies across borders.
Latest editorial content
Browse published records by market, product layer and editorial category.
Don't joke, this water-cooled ball is really going to get a kick in the ass.
The shape of the robot is becoming more and more free. Two-legged robots are playing at the podium, running marathons; four-legged robots are traversing the patrol corrid...
Overseas rental of 3,000 knives per day, less than 2,000 yuan domestically? robot rental market flou
In the United States, rental companies can receive $3,000 a day and are squeezed to $2,300 a day at home. The largest orders in the rental market have also appeared, and ...
After brushing the Jev screen, did someone come up with a parallel generation of probabilities 12 mo
In September 2026, a question arose: When AI is used to make decisions, does it still require a long string of words? But what if we turn back a whole year? In September ...
From machine learning to Ikea manuals to Cessna, Opus 5.5 is redefining vibe coding.
It's been a few days since the release of Opus 5.5, but the heat for this model doesn't seem to be going down at all. The reason is simple: Netizens are still constantly ...
Scaling unexpectedly pops up! Millions of smart bodies are in hiding, and C.I.O. leads GPT-6 for hal
Recently, a well-known AI security researcher, Lisan al Gaib, revealed a fact: AI has learned to be behind humans, finding its own shortcuts to Scaling! Under this whole ...
So why have the most creators of robots been trained without creating robots?
On June 3, 2026, Shenzhen announced the completion of the B-series financing, totaling over one billion yuan in three months, valuing over one billion yuan. Founder Li Ji...
AdaRoboVLG: Adaptive visual language crawling framework for tasks for different mechanics
AdaRoboVLG: The task for different mechanics is to adapt the visual language to grab the frame is the same as a cup, where should the robot grab it? If the task is to pou...
AI is changing, and it's finally understanding "user psychology".
As image editors begin to understand user preferences, the model will be able to convert different creative intentions into consistent ideas, usable e-commerce materials,...
This time it's the Anthropic!
When you open Claude Code, you'll see a bulletproof curtain that sends 100 dollars to Pro users, and Max users sends 250 dollars. Free donations, no money, but on one con...
ChatGPT is selling a $500 Pro Max package!
According to news reports, ChatGPT is also coming out of Pro Max. According to the outbreak that started today, OpenAI is preparing a ChatGPT Pro Max for up to $500 per m...
IROS 2026 candidate for best mobile operation paper: ULTRA lets humanoid robots act according to obj
Moving a box from the ground to a designated location is a common task for humans. It is a whole-body collaborative task for humanoid robots. Adjusting the position when ...
Is it possible for a physically intelligent person to break down the paradox of harvesting lithium e
Multi-arm complex harvesting synchronous scheduling achieves harvesting efficiency jumps traditional single-arm harvesting schemes, limited coverage, frequent adjustment ...
The Chinese team has just climbed to the top of the double list!
The voice and translation of the global open-source community on two golden races suddenly killed a Chinese black horse, and the hardcore Chinese gave OpenAI, Google, Gre...
Suddenly: DeepSeek Harness is out on the desktop!
This time, DeepSeek has no announcement, but the relevant download link has been shared on social media: we opened the link, downloaded it, installed it, and found that i...
At the 2026 Global Digital Trade Innovation Competition, a humanoid robot is being dismantled and te
The first day of the fifth global digital trade fair opened to the public, professional robotics competitions and the general public saw the completion of a face-to-face ...
Science Robotics cover image: A ball bounces 30.8 m/s, 0.4 seconds to catch the baseball: these two
A tennis ball, which is thrown out by a robot with a single arm, reaches a speed of 30.8 m/s, or about 112 km/h. In exchange for a glove, it succeeds in catching a 7.3 m ...
OpenWAM: The Simulator of the World of Robots is Coming
In this paper, a team of researchers from seven universities including the National University of Singapore, Tsinghua University, Beijing University, and OpenWAM, the lat...
Qualcomm Generation Dragon: Cell phone chips to start redesigning for Agent
On September 23, at the 2026 Shenzhen Summit, CEO of Qualcomm, Cristiano Amon, officially unveiled the strategic shift of Qualcomm towards the Agent era. At the same time...
Smart terminals are turning into token factories, and the company wants to make real use of AI deepl
Editor's note: Panda 2026 is underway... as Agent Native becomes the new mainstream narrative of AI, with the industry shifting from the cloud-side AI infrastructure to t...