Agility says every Digit humanoid is its own secured system, with a unique cryptographic identity built into the hardware, so no two units share an identity. According to CTO Pras Velagapudi, the design keeps a fault contained inside a single unit instead of letting it spread across the fleet.

D-Robotics has spotlighted WM-LOCO, a framework that pairs world models with locomotion learning so humanoid robots can predict what their actions will bring and keep their footing on challenging terrain. The work appears in the latest installment of the company's See Intelligence Happen video series. D-Robotics algorithm engineer Yuxi Liu will dig into the technology at a webinar on Nov 19 at 13:00 UTC, and the company invites viewers to watch the video and bring their questions.

Mech-Mind Robotics says generalization, the ability to handle diverse and complex objects, is a key benchmark for Physical AI, and transparent items such as plastic packaging, films and glass put it to the test. Light transmission and refraction make reliable 3D perception difficult, while variations in shape and material add complexity to robotic grasping. The company's embodied AI Eye-Brain-Hand system combines high-precision 3D vision, AI-powered understanding and intelligent path planning.

Unitree founder Wang Xingxing says robots will push computing demand to levels beyond current imagination, highlighting a potential new infrastructure strain from embodied AI.
Agility's Chief Robot Officer Jonathan Hurst calls perceiving, reasoning and doing useful work in environments a robot has never encountered before the grand challenge of robotics. Agility says its AI stack is built to solve exactly that, and points to a video for demonstration. No performance metrics or deployment figures were disclosed in the post.

ROBOTERA builds the two most expensive modules of a humanoid robot as well as the machine. Two of its five named outside buyers also build robots, and revenue is never disclosed.

Huachangda (300278.SZ) said it has deployed physical-AI capabilities for autonomous mobile robot path planning and WMS+WCS warehouse scheduling, though commercial applications remain in market expansion.
World models can forecast future states, yet robots still struggle to move well in the real world. Compounding prediction errors and the mismatch between generated actions and actual dynamics explain much of the gap.
BioMARS 2.0 is an embodied AI system that connects paper reading, experiment planning and robotic execution, allowing robots to help choose the next research step.
Robot Era has deployed its 42nd-generation embodied model in logistics, where five-finger dexterous-hand robots at China Post and SF Express sort more than 1,500 parcels per hour. The company, which raised 1 billion yuan in July, develops robot brains, motion control, humanoid bodies and dexterous hands.
Haibaichuan Technology’s humanoid robot placed second in free gymnastics at the 2025 World Humanoid Robot Games, demonstrating agile running and balance. CEO Sun Bo says small firms treat hardware development like research, and Haibaichuan is exploring detachable face accessories to make robots less intimidating.