Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
XPENG commissioned its humanoid robot production lines, and its IRON robot autonomously walked off them after completing assembly, a step the company frames as moving from R&D prototypes to real line manufacturing.
Chinese AI chipmakers, including Huawei and Cambricon, have sharply raised prices for current and next-gen processors.
Anthem is a long-range precision strike weapon aimed at a gap Covenant argues current Western arsenals lack.
Analog Devices agreed to buy Alif Semiconductor for $1.35 billion cash plus up to $200 million contingent on performance, adding microcontrollers built to run neural networks directly on-device rather than sending data to a data center.
JD Logistics plans to procure 3 million robots, 1 million autonomous vehicles, and 100,000 delivery drones over five years to automate picking, sorting, transport, and last-mile delivery.
Vention opened a Montreal lab focused on robotic manipulation, industrial data collection, and post-training robotics foundation models using learning-from-demonstration, reinforcement learning, vision models, and motion planning.
Chinese regulators are tightening approval for humanoid robotics startups seeking to go public, following a volatile trading debut from sector leader Unitree Robotics, as they try to cool a sector described as sizzling but rife with copycats and little real technological breakthrough. The China Securities Regulatory Commission's tighter posture is expected to affect other humanoid IPO hopefuls, reportedly including names such as AGIBOT and Galbot.
Robocurve, a public benefit company backed by YC, has opened a community bounty program to build reproducible open-source evaluation benchmarks for physical AI and autonomous robot systems. To make these benchmarks happen, it is giving away $20,000 and a pair of I2RT ABC Box robot arms to 25 teams who propose the best research benchmarks. Robocurve hopes to accelerate the development of Physical AI by increasing the number and diversity of dexterous robotics benchmarks.
This open-source repository presents a comparative benchmark evaluating four popular LiDAR odometry and SLAM algorithms—KISS-ICP, DLIO, FAST-LIO2, and GLIM—on live aerial drone flight data collected with a Livox Mid-360 LiDAR and Pixhawk flight controller. Running directly on a Jetson Orin NX embedded computing platform, the project compares trajectory accuracy and map reconstruction quality against millimeter-grade VICON ground-truth data. These empirical findings provide developers and perception engineers with actionable performance data to choose and tune state estimation algorithms for compute-constrained autonomous platforms.
Xiaomi's 18 Fold launched in China a full 24 hours ahead of Apple's rumored iPhone Fold, packing a wide-style, passport-sized chassis with a 5.38-inch outer OLED and 7.58-inch inner display, both hitting 4,000 nits peak brightness with a 1-120Hz LTPO refresh rate and, notably, an RGB sub-pixel structure that's a first for any foldable. Xiaomi prioritized battery over thinness, fitting a 6,000mAh cell (with 67W wired charging) into a 5.02mm-unfolded body that's still slightly thicker than Samsung's Galaxy Z Fold8.
Data collection for Physical AI is an expensive problem blocking the development of more advanced models. To get around this bottleneck, researchers at Rice University have developed RoboTok, an automated data engine that indexes, filters, and processes large-scale human demonstration videos from the internet to train manipulation policies. By extracting hand trajectories and object interactions directly from uncurated web video, the system learns complex, multi-stage manipulation skills without requiring physical teleoperation. This approach showcases how massive online video datasets can be converted into scalable training data to overcome physical data collection bottlenecks.
Orion is an open-source, 12-degree-of-freedom quadruped robot designed, fabricated, and programmed from scratch using ROS 2, custom 3D-printed components, and serial bus servos. The repository provides complete mechanical CAD models, kinematic calculations for dynamic walking gaits, and step-by-step firmware integration guides. This project provides engineering teams and builders with an accessible reference architecture for prototyping low-cost legged locomotion platforms.
They say any sufficiently advanced technology is indistinguishable from magic. Sean Tsai, an engineer at Foxconn, has built an interactive robot that performs classic sleight-of-hand magic tricks for human audiences. It integrates computer vision to track spectator choices and low-latency motor control to execute precise, synchronized mechanical deceptions in real time. It illustrates how combining high-speed visual tracking with fine motion planning can significantly enhance interactive entertainment and human-machine collaboration.
NASA is sending a seven-axis robot into low-Earth orbit, and they are looking for researchers who would like to test their code in the challenging environment of microgravity. The NASA Fly Foundational Robots (FFR) Mission is a competition where teams develop autonomous software to show off the potential benefits and challenges of space robotics. Do you think you have what it takes? The last day to submit an application is September 23.
Agility Robotics, one of the early leaders in humanoid robotics, reported $1.8 million in revenue alongside a net loss of $114.7 million for 2025 as it prepares to go public via a SPAC merger.
IonQ published a 70-page resource estimate concluding that a trapped-ion quantum computer with 19,397 physical qubits, built on its “Walking Cat” architecture, could compute a 256-bit elliptic-curve discrete logarithm on secp256k1, the curve securing Bitcoin and Ethereum, in about 25.7 days per attempt.
Huawei has become the de facto coordinator of China's push to build domestic lithography machines.
Quantinuum secured $100 million in CHIPS and Science Act funding, the only trapped-ion architecture selected in this funding round, to tackle a core scaling problem: the specialized ion traps, control electronics, lasers, and optics behind its quantum processors get harder to manufacture consistently as systems grow larger.
A New York Times investigation traces how Inspur Group, the Shandong-based server manufacturer Commerce placed on its Entity List in March 2023 specifically to cut it off from restricted US technology, kept acquiring Nvidia's most advanced AI chips for years afterward. The workaround was structural rather than clandestine: Inspur and entities connected to it routed purchase orders through affiliated subsidiaries and third-party resellers whose names never appeared on the blacklist itself, letting racks of cutting-edge hardware flow through channels the Entity List was never built to catch.
DeepSeek is preparing a large order of Huawei AI chips to power a reported one-gigawatt data center in Ulanqab, Inner Mongolia, which could begin partial operations in late 2027 or early 2028. Nvidia's most capable accelerators can't legally sell into China, leaving Huawei's Ascend family as one of the few domestic options at this scale. DeepSeek has already been tuning models for Ascend, its V4 model launched in April built for the Ascend 950, prompting Chinese cloud providers to order hundreds of thousands of units. Nvidia's Jensen Huang has called this shift a “horrible outcome” for the US, since every model optimized for Ascend chips away at CUDA's ecosystem advantage.
The Roomba Duo is a dual-robot architecture where the larger autonomous machine physically carries and deploys an ultra-compact companion using a top-mounted mechanical arm. Both robots share a map and coordinate in real time, while dual-line LiDAR, PrecisionVision AI, and additional top-mounted cameras provide navigation and perception. The main robot adds a 60 mm roller mop with 46 N downforce, steam pre-treatment, 17 cm hub-motor wheels, and a centralized autonomous maintenance dock.
This open-source repository provides a simulation-first autonomy stack for an indoor person-following quadcopter using ROS 2, ArduPilot SITL, and the Gazebo simulator. It integrates real-time visual tracking, state estimation, and dynamic trajectory planning designed to transition reliably from virtual digital twins to physical flight.
The COSMIIC System is an open-source, modular implantable neuromodulation platform designed for adoption by medical technology researchers and startups. It provides standardized, interoperable hardware modules—such as pulse generators, power supplies, and biopotential recording units networked over a Controller Area Network (CAN) bus—to accelerate translational development from benchtop testing to human clinical trials.
An attacker beaming a tuned electromagnetic carrier at a device can exploit the nonlinearity already built into ordinary components like amplifiers, ADCs, and power converters to modulate a device's low-frequency secret signals onto that carrier, upconverting them into detectable EM emissions that wouldn't otherwise radiate far enough to intercept. By tuning the injection frequency and amplitude, the resulting InjectEave attack lets an adversary actively shape how much leakage they get rather than passively waiting for it. The team demonstrated eavesdropping on audio from both wired and wireless headphones up to 30 meters away using accessible RF gear.
The SKD-1 repository provides open-source ROS 2 packages and configuration files for developing and operating a custom skid-steer mobile ground vehicle. It includes URDF robot descriptions, ros2_control hardware interfaces, and Gazebo simulation environments to support kinematic validation and autonomous navigation.
Right as Tesla's steering-wheel-less Cybercab began offering public commercial rides in Austin, NHTSA opened a formal investigation into whether the vehicle was properly certified, reviewing the technical data and compliance procedures Tesla used to self-certify up to 1,000 units under Federal Motor Vehicle Safety Standards. The core issue is hardware: the Cybercab has no steering wheel, pedals, or side-view mirrors, and automakers building vehicles without manual controls generally have to petition NHTSA for a regulatory exemption, a path Amazon's Zoox used successfully in July, but Tesla never filed one, instead self-certifying SAE Level 4 autonomy in Texas under new state rules and maintaining the Cybercab isn't subject to the standard 2,500-vehicle exemption cap.
This Thingiverse project shares a 3D-printable, three-axis manual micromanipulator designed for precise positioning in benchtop experiments and inspection tasks.
A community-maintained wiki has emerged documenting reproducible Docker builds, runbooks, and benchmarks for running frontier LLMs like GLM-5, DeepSeek-V4, and Kimi on Nvidia's RTX PRO 6000 Blackwell cards over PCIe alone, without needing NVLink.
Gravis Robotics secured $200 million in Series A funding to accelerate the commercial deployment of autonomous retrofit systems for heavy construction machinery.
Boston's Reframe Systems raised $40 million in Series A funding to expand its network of robotic microfactories that manufacture modular, zero-carbon housing components.
Uber began deploying Wayve-powered autonomous vehicles across London on Thursday, marking Wayve's first commercial public service anywhere in the world and making London only the second European city with robotaxis, after Zagreb. The initial fleet, fewer than 20 Ford Mustang Mach-Es with plans to add Nissan Leafs later, keeps a licensed safety operator behind the wheel for now, and riders booking UberX, Comfort, or Electric may get matched with one at no extra cost.
Figure signed a deal with Nscale to deploy up to 100,000 Nvidia Vera Rubin GPUs, starting in Barstow, Texas, in the second half of 2027, an initial $3.5 billion compute commitment with intent to scale past $6 billion, with Nscale also making a strategic investment in Figure. It follows last week's launch of Index, Figure's crowdsourced humanoid training dataset now generating 35 minutes of data every second, addressing the other half of the bottleneck: Figure says it's now largely constrained by data and compute rather than robot hardware itself in training its Helix model.
Micron is reportedly investigating near-GPU NAND, high-endurance flash modules placed right on the GPU package where HBM sits today, aiming to fill the massive latency and bandwidth gap between HBM and a system's SSD. The pitch is memory-bound inference: large models buy extra accelerators purely to hold weights and KV cache, an expensive way to add capacity since HBM is the priciest, tightest-supplied memory around, so a cheaper flash tier fast enough to keep compute fed could let the same model run on fewer GPUs.
Nvidia's RTX Spark superchip fuses a 20-core Grace Arm CPU with a Blackwell RTX GPU on one 3nm package, delivering up to 1 petaflop of AI performance and 128GB of unified LPDDR5X memory, enough to run 120-billion-parameter models locally with up to a 1-million-token context window. It's Nvidia's first true client SoC for Windows, pairing native CUDA support with full RTX graphics tech, so the same chip targets AI development, creative work, and gaming. Jensen Huang promises at least 100fps at 1440p in modern AAA titles.
Bimo is an open-source, 3D-printable bipedal robot kit built with custom quasi-direct drive actuators, ROS 2 integration, and simulation-ready environments for reinforcement learning locomotion policies. The repository provides complete mechanical models, electronics schematics, and training pipelines to enable reliable sim-to-real transfer on a desktop-sized robot. It offers an accessible, affordable testbed for experimenting with legged locomotion dynamics, custom motor control schemes, and modern reinforcement learning algorithms without requiring high-cost industrial hardware.
Arduino trained a generation of young engineers, and now, post-Qualcomm acquisition, they are professionalizing the framework. The recent Arduino Core on Zephyr 1.0.0 release bridges the familiar, beginner-friendly Arduino API with the industrial-grade Zephyr Real-Time Operating System. This integration allows developers to leverage enterprise capabilities like preemptive multi-threading, advanced power management, and standardized networking stacks while writing standard Arduino code.
As multi-agent AI workflows spawn dozens of independent model calls from a single task, all competing for one local GPU's execution queue, Nvidia's new open-source PAIR router spreads those calls across every eligible RTX GPU, RTX PRO workstation, DGX Spark, or Apple M4+ device already sitting on a home network. It doesn't split one inference request across multiple GPUs or pool VRAM into a bigger virtual accelerator - each request still runs whole on one node, but it proxies existing Ollama and LM Studio interfaces so agents keep using the connection they already expect while PAIR handles discovery (via mDNS), MTLS-secured pairing, and live scheduling based on which paired node is actually ready.
Dyson's CameraJet packs a 100,000-pixel macro lens camera into the brush head, running its "Gap Optical Targeting" machine learning system to detect and predict gaps between teeth in real time, triggering a jet of water within 100 milliseconds of spotting one.
TSMC Senior VP Jeffrey Hou revealed at SEMICON Taiwan 2026 that AI-driven demand has pushed the company's equipment purchase forecasts to nearly double what they were just six months earlier, as it simultaneously builds roughly 13 fabs in Taiwan and 5-6 overseas, 4-5 times its historical pace.
AGS-PlantSeg is an open-access 3D point cloud dataset and benchmark designed for semantic segmentation in complex agricultural environments across varying plant growth scales.
HTC's Vive Eagle glasses go on sale in the US, Europe, and Australia at $499 (or $599 with photochromic lenses).
Broadcom guided to $34.8 billion in fourth-quarter revenue, below the $35.1 billion analysts expected and well under some estimates that topped $36 billion, sending shares down about 4% after hours.
John Ternus, Apple's former SVP of Hardware Engineering, officially took over as CEO from Tim Cook, and used his first staff memo to hype the company's September 9 iPhone event, rumored to unveil Apple's first foldable iPhone alongside a broader Siri overhaul positioning it against ChatGPT and Claude. Ternus, a hardware-first executive taking the helm at a pivotal product moment, also teased a longer pipeline beyond next week, with camera-equipped AirPods, smart glasses, and a touchscreen laptop all reportedly in development.
CXMT has reportedly started risk production of HBM3E, a generation behind the HBM4 that Micron, Samsung, and SK hynix are already mass producing, but a meaningful jump for a company that only recently began mass-producing standard DRAM at competitive nodes. HBM is substantially harder to manufacture than ordinary DRAM, requiring high-yield die stacking, through-silicon vias, fine interconnects, thermal management, and packaging on top of competitive dies themselves, so reaching even a risk-production milestone signals China's memory ecosystem advancing past commodity DRAM.
Seraphim Space Investment Trust committed $30 million to Hubble Network, a company operating a low Earth orbit satellite constellation that connects directly to off-the-shelf Bluetooth Low Energy (BLE) chips. This architecture enables global connectivity and asset monitoring without requiring cellular towers, ground gateways, or specialized satellite modems. For builders, direct-to-satellite Bluetooth offers an ultra-low-power, cost-effective telemetry solution for tracking and managing distributed field hardware across remote regions.
Qualcomm unveiled two new Dragonwing processors aimed at bringing on-device AI, vision, and connectivity to product categories that have historically struggled with cost, power, and size constraints. The Q-2390 targets consumer and commercial gear like retail kiosks, smart appliances, home robots, and fitness equipment, integrating application compute, on-device AI, camera and display support, and LTE connectivity onto a single chip.
Atlas is an omni world model and multimodal autoregressive diffusion transformer designed to generate, reconstruct, and simulate interactive 3D environments from text, images, and video. The model natively handles spatial and camera context to synthesize novel views, reconstruct scenes into 3D Gaussian splats, and model dynamic scene interactions over time. Atlas unlocks scalable "Real-to-Sim" workflows by creating realistic synthetic sensor streams (RGB-D) and interactive physical environments from simple video captures, significantly reducing the cost and complexity of training navigation and manipulation policies.
Hydra-0 is a generalist world model that translates physical robot commands into pixel motion trajectories termed "action flow". Operating as a hybrid simulator, it pairs physics engines for robot kinematics with a learned generative video model to predict complex scene reactions like deformable materials and fluid interactions. This shared visual abstraction eliminates the need for custom coordinate remapping across different embodiments, making cross-platform policy training and sim-to-real evaluation significantly more efficient.
This open-source repository provides a simulation-first ROS 2 autonomy stack for an indoor person-following drone, utilizing Gazebo simulation and an ArduPilot Software-in-the-Loop (SITL) digital twin before physical deployment. The architecture integrates vision-based perception models for person detection with flight-control pipelines to handle autonomous tracking and indoor navigation. This project offers a clear reference implementation for designing, simulating, and validating vision-in-the-loop flight control workflows before testing on physical airframe hardware.
At Hot Chips 2026, Nvidia detailed the requirements a RISC-V CPU must meet to become a full host processor for its CUDA and NVLink Fusion platforms, alongside its existing x86 and Arm support. For CUDA, Nvidia points to RVA23 profile compliance and the RISC-V Boot and Runtime Services spec to guarantee software binary compatibility across different RISC-V chips, plus hardware-level PCIe I/O coherence and peer-to-peer GPU communication so data can move directly between host and device memory without unnecessary cache flushes or extra copies.