Atlas Builds Spot, Figure 02’s Blind Walking, AgiBot Starts Sales, WiRobot’s ALLEX

The rapid-fire developments in humanoid robotics keep coming, and the pace is only accelerating.

Atlas from Boston Dynamics is learning to autonomously assemble other robots while Figure AI’s can now get around on two feet blindfolded. And as Shanghai’s rising unicorn AgiBot starts large-scale online sales of its robots, another credible firm has entered the fray, this time from South Korea.

WiRobot’s ALLEX

The South Korean robotics firm WiRobotics is generating buzz with its new humanoid robot named ALLEX.

WiRobotics was launched by former Samsung engineers in Cheonan back in 2021. Its name stands for We Innovate Robotics. The startup initially focused on exoskeleton and wearable robotics to help workers, senior citizens, and physical rehab patients reduce strain and improve mobility.

WiRobotics' ALLEX humanoid robot platform
WiRobotics’ ALLEX (Source: WiRobotics)

They’re the latest exoskeleton maker to slide into humanoid robotics. Another buzzy startup, Paris-based Wandercraft, introduced its headless Calvin-40 humanoid robot it developed with the European automotive giant Renault Group earlier this year.

Wandercraft Calvin-40 Humanoid Robot
Wandercraft’s Calvin-40 AI-powered humanoid robot developed in collaboration with Renault Group (Source: Wandercraft/YouTube)

According to WiRobotics, the new humanoid is built as a modular platform so different parts like arms and hands can be swapped our purchased separately. For now, ALLEX lacks legs or a mobility platform, so its focus is on dexterous manipulation. The ALLEX platform integrates with WiRobot’s highly dexterous hands, which boast 15 degrees of freedom (DoF) across its fingers and wrist.

It’s intended for work across household, industrial, and healthcare settings. The name ALLEX stands for All Experience.

The humanoid is notable for its reported ability to feel touch and pressure without tactile sensors. Most robots need special sensors that simulate the senses of touch to know when they’re bumping into something. ALLEX’s developers say the robot’s joints and mechanics are so finely tuned and back-drivable that the robot can sense tiny forces through its own structure. They say the mechanics make Alex over 10 times more responsive than typical collaborative robots, also known as cobots. It can reportedly react to taps from objects as light as 100 grams or about the weight of a smartphone.

ALLEX robot hugs woman
WiRobot’s ALLEX demonstrates touch sensitivity hugging woman (Source: WiRobotics)

Despite its gentle side, ALLEX is strong enough to push with 40 Newtons of force at its fingertip, which is enough to hold tools and lift a 3 kg object with one hand. WiRobotics says the mix of extreme sensitivity and useful strength makes Alex safer and more natural to work alongside humans.

The company says it’s in active discussions with companies like Best Buy, OttoBock, and various healthcare providers as they push into international markets including the US, Europe, and Japan. By 2030, WiRobotics aims to deliver a general-purpose humanoid for everyday life that combines precision, safety, and adaptability.

Atlas Assembles Spot

In the US, Boston Dynamics just shared new demo footage to show the process its world-famous humanoid robot, Atlas, is making learning tasks autonomously.

In the video, Atlas handles and sorts parts for the leading Massachusetts robotics firm’s other famous robot, the quadruped Spot, despite an engineer’s incessant trolling. It hints at a future where AI-powered robots like Atlas assemble other robots without human oversight.

Boston Dynamics' Atlas humanoid robot assembles Spot robotic dogs
Boston Dynamics’ Atlas humanoid robot assembles parts for BD’s Spot robotic dog (Source: Boston Dynamics)

Boston Dynamics says the autonomous behavior in the demo is thanks to an end-to-end neural network it’s developing with the Toyota Research Institute (TRI). They’re building on the work TRI began back in 2023, integrating its so-called Large Behavior Model (LBM) so Atlas can respond to natural language prompts with autonomous robotic actions.

BD Atlas's digital twin trains in NVIDIA Omniverse to assemble Spot robots
Atlas’s digital twin trains to assemble Spot quadrupeds (Source: Boston Dynamics)

According to Boston Dynamics, Atlas learns through human teleoperation data. Instead of learning one task at a time, the robot trains on many of them together. The company says this allows for generalization, meaning the robot can handle new situations without extra programming. Atlas practices in the real world and within high-fidelity simulations powered by NVIDIA’s digital twin technology.

In simulation, Boston Dynamics says it can double or triple the speed it takes to learn tasks vs. real life. However, the transfer of the skills learned in simulation, a process called Sim2Real (Simulation to Reality), is never seamless.

Boston Dynamics has built a hybrid training loop to make transitions from sim to real as painless as possible. The robot’s AI brain is anchored in real-world physics since human demonstrations serve as the foundation of the LBM. Digital twins replicate all 78 DoF of Atlas and simulate how each of its joints and motors work, down to tiny forces and micro movements.

Data flywheel concept applied to humanoid robotics
Data flywheel concept as applied to BD’s robotics training (Generated with ChatGPT)

Within NVIDIA’s Omniverse, thousands of training iterations are done simultaneously. After simulation training, the physical robot runs through real-world evaluation tasks. Failures are logged, corrections are made via teleoperation, and those adjustments get back into the larger dataset. This creates what’s known as a data flywheel, where each failure teaches the AI model to make the next deployment smoother.

Boston Dynamics says it’s scaling up training, with more data, more tasks, and more diverse environments. Future iterations of the AI model could integrate touch sensing, better vision-language understanding, and learned behaviors from other robots’ experiences.

Atlas V1 in 2013
Atlas in 2013 (Source: Boston Dynamics/DARPA)

Atlas, which began as a DARPA project in 2012, is getting its learn on as it prepares for its first real job for its parent company, Hyundai. According to reports, Atlas robots have spent much of 2025 training for imminent deployment at the South Korean automaker’s US manufacturing facilities.

LG Innotek has developed next generation camera and vision components for superior perception on the factory floor. Boston Dynamics and Hyundai also announced a deeper collaboration to produce more robots.

Figure 02 Walks Blindfolded

Meanwhile in Silicon Valley, the unicorn startup Figure AI says its autonomous humanoid robots are approaching superhuman status at an accelerated rate.

Figure 02 walks autonomously in Silicon Valley
Figure 02 walks autonomous using onboard Helix AI (Source: @adcock_brett/X)

Founder Brett Adcock just shared new smartphone footage of the company’s flagship robot, the Figure 02, navigating complex outdoor terrain controlled by its new self-developed Helix AI walking controller. According to Adcock, the robot in the video relies solely on its onboard AI to move, not visual input, so it’s walking blindly in ways humans generally cannot. The advancement could make Figure’s robots useful in logistics and disaster-response scenarios where human vision could be impaired.

Figure says its Helix AI that powers industrial applications can now also accomplish dreaded domestic chores like laundry. The company has accelerated its push into households as it aims to deploy 100,000 humanoids into the world within the next four years.

AgiBot Starts Online Sales

Meanwhile in China, one of the perceived leaders in the race toward fake humans just kicked off large-scale humanoid robot sales.

Shanghai-based AgiBot has unleashed its entire lineup on JD.com, one of the world’s largest online retailers and also an AgiBot investor. Also known as Zhiyuan Innovation Shanghai Technology, AgiBot also just announced a multi-million-yuan deal with Miaoyang Fulim Precision Machining to deploy nearly 100 of its heavy-duty wheeled industrial humanoids. The unicorn startup, founded in 2023, calls it China’s first-ever large-scale industrial humanoid robotics rollout.

AgiBot on JD.com promotion
AgiBot shared this promotional image teasing its online shop’s launch on JD.com (Source: AgiBot)

On JD.com, AgiBot’s robots range in price from under $14K for the short Lingxi X2 Youth Edition and $23,500 for the Expedition A2 Youth Edition. The pricing is in line with leading Chinese robotics firm Unitree’s, whose G1 robots that start at $16K are becoming ubiquitous within and outside the mainland.

AgiBot’s online shop also includes its OmniHand dexterous robotic hand starting at just $2K. There’s also a quadrupedal bot that goes for under $5K and a commercial cleaning robot priced around $62K.

Make sure to subscribe to stay up-to-date amid rapid-fire developments in humanoid robotics. For a rundown on the Chinese humanoids that have emerged in the last few months alone, check out this related video.