Issue 1
3 – 9 Aug 2026
21 items · published, not sent

The US ban on foreign robots, and Mitsubishi joins the humanoid race

Corrected 14 Sep 2026Corrected on 11 September 2026. Six items came out: two further write-ups of the US ban on foreign-made robots, one of them published before this issue's week; a video repeating Google's Gemini Robotics 2 launch; a second write-up of the WAIC show; NVIDIA's data-centre tour, which is off our beat; and an NVIDIA explainer that was the company's third item here. Two items from the same week came in. Headlines that were repository names were rewritten, and tags and company links corrected. Later that day, the summary of NVIDIA's Video to Data dataset was rewritten from its dataset card, and four Quick Links from the same week came in: CaryX AI's everyday-manipulation dataset and three papers. On 14 September our weeks moved from Saturday–Friday to Monday–Sunday, so this issue now covers 3 – 9 August: three items from 1 – 2 August came out — Hyundai's physical-AI video, FlowPilot and Gemini Robotics ER 2 — and three from 8 – 9 August moved in from issue 2: Mitsubishi's humanoid plans, CaryX AI's RGB-D video and robot dogs doing guard work. Three items from the week came in: why VLA models fail at contact-rich tasks, a robotic-fish study and a roundup led by Unitree's IPO; the subject changed with them.
Robots & HardwareSee all ›
#
3 Aug

Sharpa's Wave hand served as verification hardware for Gemini Robotics 2 (2 minute read)

Sharpa's Wave dexterous hand, mounted on Apptronik's Apollo 2, served as verification hardware for Google DeepMind's Gemini Robotics 2. The demonstration covered multi-finger tasks including tying trash bags, unscrewing light bulbs and sealing resealable bags. Gemini Robotics 2 extends control from the upper body to the whole humanoid, feet to fingertips, and in a recorded demo guided Apollo across a room to shelve a watering can while avoiding obstacles.

#
3 Aug

When expressive humanoid robots are awkward, people become wary (3 minute read)

In a study published in Science Robotics, 50 people held conversations and made joint decisions with the commercial humanoid Pepper, which sometimes gave sound advice and sometimes interrupted or pushed illogical suggestions. When an expressive robot broke conversational norms, participants' oxytocin rose, and the higher it rose the less they trusted the robot and took its advice. Errors damaged trust whether or not the robot was expressive; expressiveness changed how their brains handled the moment.

— theconversation.comhumanoids
#
8 Aug

Mitsubishi joins the humanoid robot race, at a 1k unit per month pace (2 minute read)

Mitsubishi plans to build up to 1,000 Highlanders humanoids a month at a former engine plant in Kyoto, with mass production possibly starting next year. The robots, co-developed with a University of Tokyo startup, will work in Mitsubishi's own factories first — including building themselves at the Kyoto site. Mitsubishi is better known as a leader in automation and controls than as a carmaker, and the memorandum of understanding with Highlanders was signed last month.

— electrek.cohumanoidshardware
Models & ResearchSee all ›
#
4 Aug

Beyond VLAs: How World Action Models Reshape Robot Manipulation (5 minute read)

NVIDIA argues that a vision-language backbone learns to describe the world rather than predict how it evolves, which is exactly what a robot needs when a task depends on anticipating change. The post sets out world action models as the replacement for that backbone. NVIDIA researcher Jim Fan set out the shift in his Robotics' End Game talk, later condensed to the line that VLAs are dead.

— Robotics – NVIDIA Technical Blogvla-foundation-modelsworld-modelsmanipulation
#
3 Aug

Open-DiffLoco: Open-Source Differentiable Learning for Deployable Blind Quadruped Locomotion (paper · 15 minute read)

Open-DiffLoco is an open-source framework that trains blind quadruped locomotion policies with differentiable simulation, running the Short-Horizon Actor-Critic algorithm in MuJoCo XLA. Deployed on a Unitree Go2, the proprioceptive policy tracks omnidirectional velocity commands with under 0.2 m/s of error, runs faster than 1 m/s and withstands lateral pushes, without reference trajectories or complex auxiliary rewards. Training takes 20 to 60 minutes on a single RTX 5080 with under 6 GB of video memory.

#
3 Aug

Demystifying When and Why VLAs Fail in Contact-Rich Tasks and How to Fix Them (paper · 15 minute read)

The authors trace the failures of vision-language-action models in contact-rich manipulation to two causes: precision failures, from a mismatch in how flow-matching policies are trained, and force failures, from the structure of force signals. FACT combines a targeted fix for each and reaches a 66% average success rate on five contact-rich tasks, against 41% for the best prior baseline, over almost 2,500 real-world rollouts.

#
7 Aug

AI agents create virtual playgrounds to help robots get crucial training data (6 minute read)

A system from MIT uses three agents to assemble the objects, walls and overall appearance of a 3D scene. The generated indoor spaces let robots practise skills and try alternative approaches before they are powered on, addressing how slow and costly it is to teach machines physically. The researchers put the remaining bottleneck in content rather than physics: simulator engines have improved, but generating scenes rich and varied enough to match the real world has not.

— Robohubsimulation
Data, Sim & ToolsSee all ›
#
7 Aug

Video to Data: NVIDIA's robot dexterity task library and dataset (dataset · 1 minute read)

NVIDIA has released 1,215 human manipulation demonstrations from the TACO dataset, retargeted to bimanual Sharpa hands, with 121,500 robot episodes generated from them by its Video to Data pipeline. The 924.8 GB set, in Parquet and LeRobot v3 format under CC-BY-4.0, is meant to show what the pipeline outputs and to let users try its grounding step or downstream training.

#
3 Aug

LeRobot v0.6.1 renames a core module and loads safetensors straight to the device (tool · 7 minute read)

LeRobot v0.6.1 renames the lerobot.types module to lerobot.lerobot_types, a breaking change from v0.6.0. The release also loads safetensors directly onto the target device and improves the vision-language subtask annotation tooling. Other changes resolve policy components by convention and aggregate policy sub-losses through the metrics tracker.

#
7 Aug

Genesis 1.3.2 makes rigid-body simulation deterministic on CPU and GPU (tool · 1 minute read)

Genesis v1.3.2 makes rigid-body simulation behave consistently across scene orientations and scales, and improves numerical robustness when the mass matrix is ill-conditioned. Simulation is now fully deterministic on a given machine, on both CPU and GPU. It also adds terrain-height querying, and fixes viewer picking, raycast sensors missing hits, and a crash on delayed sensor reads.

Industry & FundingSee all ›
#
5 Aug

Trump's AI protectionism has come for robotics (3 minute read)

The FCC has banned imports of foreign-made advanced robots, including humanoids, quadrupeds and wheeled machines. The order cites the data such robots would collect in homes and sensitive facilities, and the need to shield US robotics firms from Chinese competition. The playbook predates this administration: the same tariff-and-procurement response followed Chinese solar panels, electric vehicles and drones.

— technologyreview.compolicy-safetyhumanoids
#
3 Aug

Walden Robotics Partners With Toyota on Practical Humanoids (5 minute read)

Walden Robotics, the Toyota Research Institute spinout that left stealth on 15 July with $300 million at a $1.1 billion valuation, is partnering with Toyota on humanoids built for practical work. IEEE Spectrum sets it against two years in which humanoid companies were founded largely because it was possible and lucrative, and the industry-wide sobering about practicality and value that followed.

— IEEE Spectrumhumanoidsindustry-funding
#
5 Aug

China’s X Square Robot submits confidential filing for Hong Kong IPO: sources (1 minute read)

Shenzhen-based X Square Robot has filed confidentially for a Hong Kong listing, with Huatai Securities and Morgan Stanley as sponsors. AgiBot and Galbot have filed confidentially as well, according to sources cited by the South China Morning Post. The rush comes as Hong Kong's IPO market hit a five-year high in both funds raised and number of listings over the first half of 2026.

— South China Morning Postindustry-funding
#
3 Aug

Unitree opens its STAR Market IPO to subscription on 10 August; AgiBot starts a Hong Kong listing (7 minute read)

Gasgoo's weekly roundup leads with Unitree's STAR Market listing: the preliminary inquiry was set for 5 August and subscriptions for 10 August, for 40.45 million new shares, 10% of the enlarged capital. Unitree projects first-half revenue of 1.05 billion to 1.13 billion yuan, up roughly 35.6% to 45.4% year on year, and net profit attributable to shareholders of 258 million to 306 million yuan. The roundup also notes AgiBot starting its Hong Kong listing process.