robotics

🧠 Changing the plan, not the model, lifted a robot world model's success from 22% to 85%

Researcher Haddy Alchaer ran 600 experiments on LeWM, a JEPA world model from a team that includes Yann LeCun. As published, it succeeded on 22% of simulated tasks where a robot must push a T-shaped block to a goal 75 steps away. A new subgoal every 25 steps raised that to 80%, and a memory of past demonstrations with a cost on every jump between them reached 85%. On paths the model was trained for, its own predictions scored 78%, and the simulator's real physics scored 77%. No weights were retrained.

⭐ The useful fix was a shorter plan, not a smarter predictor.

#AI #Robotics

🤖 GXO just signed for up to 1,000 humanoid robots in its warehouses

GXO, the world's largest pure-play contract logistics company, signed a letter of intent with Apptronik on November 12. From 2027, Apollo humanoids go on trial in live warehouses: picking for a person and from a person, moving cases and totes, packing, and sorting. If those trials hit the safety and output targets both sides agreed, they intend to roll out up to 1,000 Apollos worldwide over four years, with the first North American deployment in late 2027, paid as a robot service instead of a one-off purchase.

⭐ A warehouse giant is not buying a demo. The thousand robots only show up if Apollo survives a real shift.

#AI #Robotics

🚗 Tesla starts selling a car with no steering wheel: Cybercab orders open at $29,990

Tesla is inviting "select customers" to order the self-driving Cybercab for personal use at $29,990 before taxes, with a $1,000 non-refundable deposit. The two-seater has no steering wheel or pedals, and it needs a $299/month "Autonomy Subscription" to operate. It only works inside Tesla's Robotaxi zones in Texas and Florida, and first deliveries are estimated for December 2026 in Austin.

⭐ Musk's "under $30,000" robotaxi is real, but owning a car that drives itself still comes with a monthly bill and a geofence.

#AI #Robotics

🤖 OpenAI just disbanded its robotics team, because ChatGPT-style models are starting to run robots themselves

OpenAI confirmed to Fortune that it broke up its stand-alone robotics team and moved the researchers into its main research and applied groups. The reason: its general frontier models, including GPT-6 Astra, turned out to work directly on robots. In an Nvidia study, an Astra-powered agent hit 74.8% on a simulated manipulation benchmark, versus 51% for the best specialized robot models. OpenAI says it still plans to show its own humanoid in 2027.

⭐ If one general AI brain can drive robots, startups building special robot brains face a real test.

#AI #Robotics

🚐 The retro VW Bus is coming back as a robotaxi, and May Mobility wants 10,000 of them

May Mobility plans to turn Volkswagen's electric ID. Buzz into a driverless ride-hail van, targeting more than 10,000 on robotaxi platforms worldwide through 2031. Road tests in the US start in early 2027, with the first production vans due in 2028. Partner ECARX will fit May's self-driving kit into the van, and the company says the setup cuts vehicle costs by more than 50%. May has already given 600,000+ autonomous rides in the US and Japan, including with Lyft in Atlanta.

⭐ Cheaper, purpose-ready vehicles are what robotaxis need to scale beyond a few cities, and this is one of the biggest fleet targets yet.

#AI #Robotics

🤖 ASUS unveils Gallop, its first humanoid robot, already learning a real factory job

ASUS has revealed Gallop, a two-armed humanoid built on Nvidia's Jetson Thor platform that is training at an ASUS group electronics plant to pack motherboard accessories: 15 small parts, many in thin clear bags, about 20 seconds per board. It learns mostly by watching skilled workers, and the target is 99.5% success within that 20 seconds, the same bar set for people. A second "supervisor" AI checks every step and re-plans when something goes wrong. A wheeled version and an in-house touch-sensing hand come next, with customer pilots planned for 2027 and sales targeted for the second half of 2028.

⭐ One of the world's biggest PC makers is entering the humanoid race by starting with a hard, real production task instead of a staged demo.

#AI #Robotics

🤖 SoftBank and Nvidia back German humanoid maker Agile Robots at an $8.5 billion valuation

Munich-based Agile Robots has raised a $500 million Series F led by SoftBank Group, with Nvidia, Hitachi and T.Capital also joining, lifting its total funding to about $2 billion. The DLR spin-off builds humanoid and mobile robots, robot arms and hands, plus the AI models that control them, and says it has installed more than 20,000 robotic systems worldwide. The money will fund a push into the US and more production, including its new factory near Munich that can build up to 10,000 legged and wheeled humanoids a year.

⭐ It's one of Europe's biggest bets yet on "physical AI," and a sign the humanoid race isn't just a US-China contest.

#AI #Robotics

🇨🇳 Beijing wants 5 world-class humanoid robot makers and capacity for 200,000 robots a year by 2028

Beijing has released a 2026-2028 action plan for embodied AI robots, aiming to build at least five internationally leading and ten nationally leading humanoid makers by 2028. The city targets 200,000 complete robots a year of production capacity, plus new embodied-AI computing clusters of over 50,000 PFLOPS and more than 1 million hours of real-world robot training data from at least 100 sites. Robots should carry 40%+ of their own weight and run over 2,000 hours between failures.

⭐ China's capital is turning humanoids from lab demos into an industrial policy with hard production and reliability targets, raising the stakes in the global robot race.

#AI #Robotics

🤖 Runway opens up Praxis-1, a robot brain that learned from video

AI video company Runway has released Praxis-1, its "world action model" for controlling robots, with open weights under Apache 2.0, including for commercial use. The 27B-parameter model predicts what a scene will look like next and the motion to get there, and Runway says it tops Nvidia's RoboLab-120 leaderboard at 46.2% success, ahead of every open-weight rival. Fine-tuned on Google's RT-1 robot, which it never saw in training, it hit 80% success versus 52.4% for RT-1's original policy.

⭐ Runway argues web video can stand in for costly robot demonstration data, and open weights let any robot maker build on it.

#AI #Robotics

🤖 Joyson delivers a "brain backpack" that gives Unitree robots extra on-board AI power

Chinese auto-parts giant Joyson Electronics says its robotics unit delivered the first high-compute "brain backpack" to Unitree on November 10, moving their strategic partnership into product delivery. The add-on packs an on-device computing platform and its own energy system, and fits Unitree's humanoid, four-legged and wheeled robots. It is meant to boost perception, data collection, decision-making and motion control, run embodied AI models locally and extend working time. The two also plan joint work on key robot parts and factory use cases.

⭐ Bolt-on compute could let existing robots run bigger AI models without a full hardware redesign.

#AI #Robotics

🏠 Chengdu unveils its fifth-gen Konka robot, built to remember chores across days

The Chengdu Humanoid Robot Innovation Center released the fifth-generation Konka (Gongga) robot on November 11, calling its new embodied AI system its "most powerful brain." Its headline feature is a spatiotemporal memory engine that tracks where objects were placed, when they moved and how past attempts went, so it can handle household tasks that span multiple days. The center also says it can plan, simulate actions and recover from errors on its own; it has not shared pricing or independent test results.

⭐ Memory is one of the biggest gaps keeping robots out of messy real homes, and Chinese labs are racing to close it.

#AI #Robotics

🤖 China's Topstar unveils Star, a 172 cm humanoid built for real factory lines

Chinese industrial robot maker Topstar unveiled Star, a full-size bipedal humanoid, at its TOPS DAY 2026 event in Dongguan. It stands 172 cm, weighs 72 kg, lifts 10 kg per arm with 7-axis arms, and is sized for existing production lines to handle and sort materials. Topstar also signed a humanoid procurement deal with Kaihang Technology, though it has not disclosed pricing, mass production timing or first customers.

⭐ China is pushing humanoids from stage demos into factories, with a state initiative aiming for routine deployment in key scenarios by the end of 2026.

#AI #Robotics

🦾 Path Robotics raises $300M to let AI welding robots handle warping metal

Path Robotics raised a $300 million Series E led by Georgian, with the venture arms of Yamaha Motor and a major South Korean tech firm joining, at a $2.45 billion post-money valuation. Its Obsidian AI model uses lasers and cameras on the torch to watch metal parts distort under heat and adjust the weld on the fly, trained partly in simulations from a "weld world model." The company already uses it in welding cells and in Rove, a four-legged robot for shipbuilding.

⭐ Skilled welders are scarce, and AI that copes with messy real-world parts is what finally makes this kind of factory work automatable.

#AI #Robotics

🚗 Nvidia open-sources Alpamayo 2 Edge, a self-driving AI that "thinks" only when it must

Nvidia released Alpamayo 2 Edge, a 2-billion-parameter vision-language-action model that turns surround-camera video and navigation input into driving trajectories, built to run in real time on its DRIVE AGX Thor car computer. It writes out step-by-step reasoning only in tricky scenes, saving compute during routine driving, and reaches 160-190 ms from pixels to trajectory when optimized. Nvidia says closed-loop reinforcement learning in simulation lifted driving performance by 48.9%, and weights are on Hugging Face.

⭐ Shrinking reasoning models to fit inside a car is a key step for robotaxi and driver-assist makers building on Nvidia's stack.

#AI #Robotics

🤖 China's AGIBOT open-sources LIFE, so robots can keep a "heartbeat" and memory between jobs

Chinese humanoid maker AGIBOT has released LIFE, a framework for robots to keep sensing their surroundings, store retrievable memories, and reuse skills after a task ends. It pairs a lightweight decision model called RoboBeats, with open weights at 0.8B, 2B and 4B, with a Plastic Memory store and a method that turns solved problems into tools. AGIBOT says RoboBeats-2B processes a new frame in about 75 milliseconds on one RTX 5090, and that on the FindingDory benchmark, reusing learned skills raised success from 37.5% to 65.8%. The paper, code and weights are public.

⭐ Most robot stacks reset when a job ends; this aims to let machines accumulate experience more like a coworker than a one-shot demo.

#AI #Robotics

🤖 China wants a lifetime "ID card" on every robot it makes

China's industry ministry (MIIT) has published draft rules requiring every complete robot made in the country, from industrial arms to service and special-purpose robots, to get a unique, permanent and tamper-proof ID code before leaving the factory. Makers would register products, including key components and software, on a national platform and file every code they use within 20 working days. Public comments are open until December 5, and the draft sets no start date yet.

⭐ It extends a scheme that already coded 28,000+ humanoids to an industry making close to 100,000 industrial robots a month, giving Beijing a detailed view of who builds what.

#AI #Robotics

🏠 Meta quietly built a household humanoid robot with China's Fourier

Meta and Shanghai-based Fourier built a humanoid prototype codenamed Max in about three months, The Information reports. The roughly 1.5-meter, 48kg robot runs on Nvidia's Jetson Thor, has a screen face and five-fingered hands for small objects, and work on Max 2 with better hands, a projector and stair sensors began in September. Meta says the project is internal research and it has no plans to sell the robots.

⭐ Meta sees humanoids as a physical body for its Muse AI agent, and leaning on a Chinese partner shows how central China's supply chain has become to robotics.

Sources: Engadget https://www.engadget.com/2303434/meta-reportedly-built-a-humanoid-robot-prototype-that-can-do-household-chores/ | Interesting Engineering https://interestingengineering.com/ai-robotics/meta-develops-humanoid-robot

#AI #Robotics

🚨 ELON MUSK'S BIG VISION: ROBOTS VS. BANKRUPTCY! 🤖
After his experience with DOGE, Elon Musk highlighted a major challenge facing the modern economy: governments are struggling with growing debt, rising interest payments, and the difficulty of controlling spending.
But what if technology offers a way forward?
Musk's vision points toward two possible futures:
🔴 SCENARIO 1: ECONOMIC CRISIS
If productivity doesn't grow fast enough, government debt could become increasingly difficult to manage. Rising interest costs, economic stagnation, and political instability could follow.
🟢 SCENARIO 2: AI & ROBOTICS REVOLUTION
Artificial intelligence and robots could transform the global economy:
Higher productivity.
Lower production costs.
Cheaper goods and services.
Faster economic growth.
Potentially greater prosperity.
🔥 MY TAKE:
The real opportunity isn't just about building smarter robots. It's about creating an economy where technology improves everyday life and makes prosperity more accessible.
However, technological abundance isn't guaranteed. The benefits must reach ordinary people, not just large corporations.
🚀 The next major economic race could be ROBOTS VS. BANKRUPTCY — technological abundance or economic stagnation?
💬 Which future do you think is more likely?
#ElonMusk #AI #Robotics #DOGE
CLICK TO BELOW TRADE👇
$FET $TAO $RENDER

🤖 The Next AI Revolution Won’t Stay on Your Screen.

For years, the world was fascinated by robots that could walk, dance and perform impressive movements.

But the real technological race is changing.

China is now shifting its focus toward Embodied AI — technology that enables robots to understand their environment, learn new tasks and make decisions in the physical world.

The numbers tell the story:
$13 billion invested.
322 deals.
Five times the investment compared to the previous year.

But here’s the interesting part:

The next breakthrough won’t necessarily belong to the company building the most impressive robot.

It may belong to the company developing the intelligence behind it.

Hardware gives a robot a body.
AI gives it a brain.
Data gives it experience.

And trust determines whether we can actually use it.

The future isn’t about making machines look human. It’s about making technology genuinely useful to humans.

#Aİ #Robotics #artificialintelligence #tech