Amazons GEN 3.5 AI Robot Launch (AI NEWS)

And this human form factor allows it to pack what’s most important, its power. So the 4NE1 can lift up to 100 kilograms or 220 pounds. Then, for speed, it moves at 5 kilometers per hour, or just over 3 miles per hour, and runs on hot swappable batteries to enable between 6 to 8 hours of continuous operation, powering its total of 55 degrees of freedom across the body. On top of this, Neuro also developed a patent-pending artificial skin that detects contact just before it happens, as well as a touchless sensor that distinguishes between humans and objects, even through obstructions, and the robot itself improves by learning autonomously through interactions and then updates its behavior over time.

But individual intelligence is only half of this equation, because the real skills come from what happens across the entire fleet, as Neuro built a shared intelligence platform called the Neuroverse, so that when 1NE1 robot learns a task, that knowledge propagates across every connected unit instantly. But how much does it cost? So the full-sized Gen 3.5 is priced at roughly 98,000 euros, with that price dropping to around 60,000 euros when purchasing a fleet of robots, and one of the largest companies in the world just signed on for this. As this week, Neuro just announced a collaboration with Amazon Web Services and Amazon, covering four areas.

First, AWS will serve Neuro’s primary cloud providers and host the Neuroverse platform for physical AI training, real-time data processing, and shared intelligence across its robot fleets. Second, Neuro’s training facilities will integrate with AWS services, including Amazon SageMaker to accelerate AI training pipelines that combine real-world sensor data with simulation. Third, Neuro will collaborate through the AWS partner network to expand go-to-market activities for its cognitive robotic solutions. And fourth, Amazon will explore deploying the 4NE1 in select fulfillment centers, providing real-world data for one of the most demanding logistics environments on Earth to make every unit smarter.

But the remaining question here is, would you purchase one of these robots for your home, or would you rather wait for it to expand its abilities and possibly lower its price tag? And next we have Unitree’s newest version of its G1, with wheels. This starts at $13,500. And tell me down below if this would be a legitimate replacement of either a police officer or a delivery driver, or possibly something else. It’s imaginable that this could chase down a human and ram into them to incapacitate them. Next we have this robot on skates.

The first thing that comes to mind here is the X Games. What would happen if you launched this off of a ramp, spinning and flipping? And finally, ice skating, which if this had dexterous hands to hold a hockey stick, it would be interesting to see a 1v1 hockey match between a human and this robot to see who would But Nura is only one German robotics company that’s preparing for prime time, because at Hannover Messe 2026, Europe’s largest industrial trade show, Munich-based Agil Robots just revealed its Agil-1, a humanoid built specifically for manufacturing.

It stands 174 cm tall, or just over 5’8″, and weighs 69 kg, or about 152 lbs. And it walks at a speed of 2 meters per second, or about 4.5 miles per hour, which is fast enough to keep pace with human workers moving between stations. It also carries payloads of up to 20 kg, or 44 lbs, and runs for up to 8 hours on a single battery charge. Plus, there’s a smaller version called the Agil-1S for research and development. But what sets Agil-1 apart isn’t its size or speed, but its hands. Because each of Agil-1’s 5-fingered hands has 21 degrees of freedom with force torque sensors and tactile feedback in every single joint.

In fact, the robot has 71 degrees of freedom total throughout its body, and at the show, it demonstrated this by picking up individual components and assembling miniature robotic arms live on stage, which requires both precision and firm grip strength in the same sequence. But what makes these hands useful beyond a demo is what’s driving them. Because the Agil-1 runs on what Agil-Robots calls its robotic foundation models, trained on one of Europe’s largest real-world industrial datasets combined with simulation and human teleoperation data. And what’s more is that its AI uses a layered architecture, where one level handles strategic reasoning and task planning, another manages rapid responses, and a third controls fine motor precision.

And for perceiving its environment, it uses a combination of cameras, LiDAR, and proximity sensors, plus it can even communicate with human workers using voice interaction, expressive eyes, and its chest-mounted display. And as for its strategic partnerships, it’s backed by both Google DeepMind and Deutsche Telekom’s Industrial AI Cloud, which is Europe’s first cloud infrastructure built specifically for industrial AI. And Agil-Robots already has over 20,000 automation solutions deployed across automotive, consumer electronics, healthcare, and logistics worldwide, supported by more than 1,500 R&D specialists. So the Agil-1 humanoid isn’t just a concept, it’s actually being manufactured in-house at a facility in Bavaria right now, designed to integrate directly into Agil-Robots’ existing ecosystem of robotic arms, mobile robots, and its Agil-Core software platform.

But beyond robots learning to see the physical world, AI is also learning to see your design system, as Google Stitch just open sourced its design MD format, which is a file that lets you export your design rules from project to project, so AI tools understand the reasoning behind your design system. So instead of guessing intent, AI agents can now read exactly what a color is for, and validate choices against WCAG accessibility rules, and this spec is available on GitHub, allowing anyone to generate their own files directly in Stitch right now. And speaking of AI that writes code, there’s also a brand new open source model that just came out.

Moonshot AI just released Kimi-K 2.6, which is an open source model built around long horizon coding, agentic execution, and something called agent swarms. And on benchmarks, it’s competing directly with GPT 5.4, Claude Opus 4.6, and Gemini 3.1 Pro across coding, reasoning, and agentic tasks. And in several categories, it’s even outperforming them. So on Software Engineers Bench Pro, which tests real world software engineering, K2.6 scored a 58.6, which is ahead of GPT 5.4 at 57.7, and Claude Opus 4.6 at 53.4. Anyways, like and subscribe. Check out this video here for more on the latest on AI and robotics news.

And thanks for watching. [tr:trw].