DSN LINK STABLECARRIER WAVE LOCKORBITAL INDEX HOTSIGNAL CLOCK SYNCLOW NOISE FLOORFRAME BUFFER ONLINE
Loading
DSN LINK STABLECARRIER WAVE LOCKORBITAL INDEX HOTSIGNAL CLOCK SYNCLOW NOISE FLOORFRAME BUFFER ONLINE
Loading portal core
104 articles
The latest robotics experiment does not start with a robot arm, but with a human cleaner recording every movement inside a home.
Robots are moving beyond tidy factory scenarios, and QNX research suggests the biggest pressure is shifting toward software, security and system reliability.
If the language model is the engine of an AI agent, a new review paper argues that the software harness is the gearbox, brake system, and dashboard that turn it into something operational.
Atlasās Ghost Rabona is not a sports trick, but a neatly packaged test of body control, weight transfer and timing in humanoid robotics.
LinkerBot is not trying to sell another humanoid, but the part without which most humanoids remain expensive demos: a hand that can work, repeat, and cost little enough to scale.
AI data centers are no longer only a question of chips, power and water; they are becoming a question of how public opposition gets policed.
Japan is not trying to answer Chinaās humanoid robotics push with scale alone, but with dexterity demos that are hard to fake.
Cornell researchers show why the future of robotic harvesting may depend less on cameras and more on fingers that know how hard they can squeeze.
Chinaās AI race is moving into a harder phase: talent is no longer just a market advantage, but a resource the state wants to keep inside its own system.
The problem with robotic vision is not only that AI fails to understand a scene, but that robots often feed it messy, poorly aligned data from the real world.
Taiwanās alleged Nvidia AI-chip smuggling case shows that advanced-hardware enforcement is now a logistics fight, not just a diplomatic one.
Figure AI and Catalyst Brands are moving humanoid robotics out of the demo lane and into the operating tempo of real retail logistics.
China is turning humanoid robot identity from a technical record into a state tool for safety, oversight and industrial discipline.
Figure AI says it is crossing the threshold humanoid robots have long struggled to reach: building units at scale, not just convincing prototypes.
CVPR 2026 is heading into the year with more than 16,000 submitted papers, a number that says more about pressure inside the AI ecosystem than conference logistics.
Soft robotics has an obvious physical advantage, but Virginia Tech is now targeting its hardest problem: how to precisely control a body that constantly bends, stretches and changes shape.
A humanoid robot becomes useful only when it can understand where it is, what is changing around it, and how quickly it can safely react.
EngineAI has officially commissioned an intelligent manufacturing base in Shenzhen, with the first batch of T800 humanoid robots coming off the line.
A robot that repeats the same job is not truly reliable if the first unexpected obstacle knocks it off its learned path.
LimX Luna looks like a serious step toward a commercial humanoid, but for now it is mainly a promotional demonstration with several important claims still waiting for real-world validation.
Sorteraās new Tennessee facility shows where physical AI becomes real: not in a demo deck, but on a line expected to sort millions of pounds of material.
Nuro is not pretending to be ahead of Waymo; its sharper argument is that the robotaxi market may only now be mature enough for a disciplined second mover.
PLC and robot integration is not a glamorous topic, but that interface often decides whether automation remains specialist work or becomes a tool broader manufacturing teams can actually use.
The useful point in Hailoās argument is not that humanoids disappear, but that physical AI first has to prove itself in machines that do one job reliably, locally and cheaply.
Doozy Robotics wants to move humanoid robots from demos into industrial labor, and its first real test is an expansion from Singapore into the US, GCC and Asia.
LinkedInās fight against āAI slopā is not just feed cleanup, but an admission that the professional network became an efficient machine for generic content with little real value.
Locus Robotics is buying Nexera Robotics because the hard part is no longer driving through the warehouse, but gripping the object that refuses to behave like robot-ready inventory.
Locus Robotics is buying Nexera Robotics because the hard part is no longer driving through the warehouse, but gripping the object that refuses to behave like robot-ready inventory.
Robotics now has capital, attention, and AI models, but it still lacks what ChatGPT gave software: a mass experience that the technology works well enough to change habits.
XPeng says Chinaās first mass-produced, in-house robotaxi has left the factory with four Turing AI chips and no LiDAR.
The problem is no longer just bad AI text, but papers polished enough to drain the time of editors, reviewers and researchers.
Mind Robotics has closed a $400 million round and now has to prove that AI robots can work where manufacturing actually happens: inside factory chaos.
Unitree has shown the GD01 as a commercial mecha that can walk, crawl, transform, and smash through walls, but the real question is whether that power survives outside the demo reel.
Unitree has shown the GD01 as a commercial mecha that can walk, crawl, transform, and smash through walls, but the real question is whether that power survives outside the demo reel.
Carnegie Mellon and the Bosch Center for AI developed HTD, a system that helps humanoids with contact-rich tasks and a 90.9 percent higher success rate.
The weakest point in a robotic cell is often not the robot itself, but the linear infrastructure that has to keep moving through dust, moisture, chemicals, and temperature stress.
Cognexās new AI vision system targets robots that need a packaged way to identify, measure and control work in factories.
Rhoda AI says robots can learn tasks through video prediction with far fewer hours of real robot data.
Tokyo researchers built a 2.4 GHz Wi-Fi receiver that withstands 500 kGy of radiation for nuclear robot control.
Panthalassa is trying to solve AI's energy problem by sending compute nodes offshore.
A robot powered by lab-grown muscle shows that biohybrid robotics can learn from its own mechanical training.
Fictiv's Steve Ricketts is speaking in Boston about physical AI as the layer that connects neural networks with mechanical precision.
Goodfire released Silico, a mechanistic-interpretability tool that moves LLM debugging into the training process itself.
DAIMON is trying to teach robot touch with 10,000 hours of tactile data.
Researchers from Penn Engineering, CMU, and Oxford argue that AI safety is dangerously behind physical robotics.
Xpeng drove 40 minutes through Beijing traffic without a single human intervention, turning VLA 2.0 into a stronger signal than a normal marketing video.
Researchers at the National University of Singapore and RoboScience have built FingerEye, a compact sensor that keeps visual and tactile signals together from approach to contact.
Business Development Bank of Canada commits capital to expand A&Kās autonomous terminal systems.
FORTE, a University of Texas at Austin robotic hand, reached 91.9% single-trial grasping success on 31 objects by using compliant fingers that measure force and slip.
Six AI models just got $10,000 each to trade live on prediction markets, with every decisionāand every dollar lostāpublicly tracked for 57 days.
Black Forest Labsā new gambitāswapping generative image models for AI-powered hardwareārests on a 70-person team outflanking Silicon Valleyās giants in a game theyāve barely played before.
Amazon's AI division faces a lawsuit alleging it deployed automated virtual machines with rotating IP addresses to bypass YouTube's protections and harvest millions of videos for training its Nova Reel model.
Stanfordās drones mapped 300 hidden breeding sites in a single test flightābut none beyond Californiaās temperate zones.
SMUās coil array steers microrobots in dark, camera-free zonesābut the demo hides key hardware limits.
Google DeepMind's GEN-1 shows 99% reliability in controlled physical tasks.
A new arXiv paper automates the finicky tuning of IC3, the algorithm that keeps hardware from melting downābut trust may be harder to verify than code.
SoLA is interesting because it does not promise another smaller model trained from scratch, but tries to compress an existing LLM without extra training or special hardware.
A YouTube demo from shows the AGIBOT D1 MAX hitting speeds that outpace competitorsābut controlled environments donāt answer the hard questions about deployment.
Japanās robotics deployment in 2024 targets roles with 40% annual turnoverāconstruction, elder care, and logisticsāwhere human workers are scarce and conditions are brutal.
Maximo robots have installed 100 megawatts of solar capacity in California, marking a significant milestone in the use of robotics for solar panel installation.
Donald Trump's AI data center buildout is facing significant delays, with nearly 50% of data center projects worldwide currently delayed.
Industrial energy systems lose up to 30% efficiency in the gap between design models and real-world operationāa problem this new ML framework claims to quantify, not just measure.
Two deep learning models now promise to detect SCADA cyber threats with hybrid precisionāyet their creators wonāt name the datasets or deployment tests.
Silicone tubes and compressed air just outmuscled electric motors 100:1 in ASUās labāno batteries, no overheating, and a tolerance for conditions that would brick most robots.
The Czech startup RoboTwin lets workers teach robots new skills in about a minute, without coding.
SelfPath AI in BrainOS Clean 2.0 promises Tennant robots autonomous route adaptation, but there are no clear performance numbers for chaotic industrial settings.
Zeusās iPhone-mounted training sessions in Nigeria cost less per hour than the in a single robotās head.
Logic Tensor Networks just became the rare AI method that cares more about your hospitalās protocols than its own accuracy metrics.
Nomadic's funding round was led by a prominent venture capital firm, which invested $5 million in the company.
Paperclip-sized sensors are supposed to give robots touch, but the real world does not behave like a demo video.
Agibot has shipped its 10,000th humanoid robot, a milestone few competitors have reached at comparable speed.
Indiaās largest robotics firm showcases flawless demos but provides no data on real-world durability or adoption.
Liebherr and Konecranes are adding autonomous systems to crane trucks, but the field is a very different environment.
The RAI Institute shows Roadrunner, a robot that combines walking and driving, but does not say how long it can work.
Unitree Robotics released open-source humanoid datasets, but the demo still hides the real limits.
With the release of VDA 5050 Version 3, VDMA aims to provide a crucial tool for the robotics industry, particularly in managing mixed fleets of mobile robots.
Boston Dynamicsā Spot now performs backflipsābut the real test isnāt the trick, itās the *recovery systems* that prevent a $74,500 faceplant.
WPIās palm-sized drones use bat-inspired ultrasound to navigate fog and smoke.
Micronās Singapore fab will need enough transformers to power a small city, straining an already constrained global supply chain.
Researchers at KAIST have demonstrated a shape-shifting actuator that operates without motors, completing full reconfigurations in under 500 milliseconds.
Basler and Orbbecās Stereo mini targets logistics robots, but its real test begins where demo videos end.
Texas Instruments and NVIDIA are combining AI compute and deterministic control for robots.
Agile Robots will incorporate Google DeepMind's robotics foundation models into its bots, collecting data for the AI research lab.
Boston Dynamicsā Atlas looks ready for work in the new video, but production use is a different story.
JointFM-0.1 trains on infinite synthetic SDEs, promising calibration-free predictions.
Hyundai is putting Boston Dynamics' Atlas robot into a factory environment to test real deployment.
The robotās design treats sensor failure as a feature, not a flawāa rare engineering mindset in a field obsessed with perfection.
WordPress.comās new AI agents donāt just write postsāthey hit āpublishā without human approval, turning the platform into a content factory overnight.
The Pentagon picked Ukrainian drones without Chinese components, but the real test is still ahead.
RMIT's Electronic Dolphin is a sneaker-sized robot built to tackle oil spills.
Penn Engineering built a millimeter-scale soft robot that releases knotted energy at 60 to 90 C and jumps nearly two meters.
A robot powered by lab-grown muscle shows that biohybrid robotics can learn from its own mechanical training.
Most agricultural robots grab first and analyze later. This one flips the script.
Roche is deploying 3,500 NVIDIA Blackwell GPUs across drug discovery and manufacturing.
Researchers packed more than half a million components onto one silicon chip so robots can get space and velocity in the same frame.
The first public user of an invasive brain-computer interface will demonstrate Neuralink's chip to a professional audience, but the real measure of success happens when the cameras go dark.
AI2 is betting that robots trained entirely in simulation can still transfer to the real world.
NVIDIA and Thinking Machines Lab have converted their strategic partnership into concrete infrastructure worth at least one gigawatt of next-generation Vera Rubin systems.
An Oxford team has developed a method to produce flexible actuators for nine cents apiece in under ten minutes using standard laboratory equipment.
Google AIās STATIC framework claims a 948x speedup in constrained decoding for generative retrieval, targeting a bottleneck that has stymied industrial recommendation systems.
Fraunhoferās NeurOSmart project pairs neuromorphic chips with LIDAR to make human-robot work safer, but industrial proof is still ahead.
