China’s PNDbotics humanoid learns in hours without setup

PLUS: Schaeffler deploys hundreds of humanoids, X Square raises $140M, and Maize Runner robots


China’s PNDbotics humanoid learns in hours without setup

Welcome back to your Robot Briefing

A Chinese startup claims its humanoid can master new tasks in hours, not months. PNDbotics' Adam-U Ultra ships with a pre-trained vision-language-action model and 10,000 real-world scenarios already loaded, eliminating the setup and calibration phases that typically stretch deployment timelines.

If robots can truly go from unboxing to productive work in days instead of quarters, the business case for automation changes overnight. The question is whether these pre-trained models deliver on their promise in messy, real-world conditions, or if we're still years away from plug-and-play humanoids.

In today's Robot update:

PNDbotics humanoid learns tasks in hours without calibration
Schaeffler commits to hundreds of humanoids by 2027
X Square raises $140M for physical AI foundation models
Autonomous farm robots save $45K while cutting chemical use 70%
News

China's PNDbotics unveils humanoid that learns in hours

Snapshot: Chinese robotics firm PNDbotics released Adam-U Ultra, a humanoid platform that masters complex manipulation tasks in hours without setup or calibration. The robot comes pre-loaded with a Vision-Language-Action model trained on 10,000+ real-world scenarios, potentially slashing the time and cost of deploying humanoids for practical work.

Breakdown:

The platform ships with a pre-trained VLA model and over 10,000 real-world data samples already loaded, eliminating the configuration and calibration phases that typically add weeks or months to robot deployment timelines.
PNDbotics claims the robot can acquire and deploy new skills within hours by training on minimal task-specific data, compared to traditional approaches that require extensive programming and testing cycles.
The company positions Adam-U Ultra as a unified platform for industrial, commercial, and research applications, featuring quasi-direct-drive joints for precision control and a modular design that supports up to 44 degrees of freedom.

Takeaway: If pre-trained models can truly compress deployment timelines from months to weeks, they fundamentally change the ROI equation for companies evaluating humanoids. The shift from "build and train from scratch" to "download and customize" mirrors what happened with large language models, potentially accelerating mainstream adoption by removing a major technical barrier.

News

Schaeffler to deploy 'hundreds' of humanoids in factories

Snapshot: German industrial giant Schaeffler has signed a strategic partnership with SKL Robotics (Humanoid) to integrate hundreds of humanoid robots into its global production network by 2027.

Breakdown:

The partnership works both ways: Schaeffler supplies actuators to Humanoid for their robots while simultaneously buying those robots for deployment across its own factories, creating a vertically integrated relationship that reduces risk for both companies.
Initial deployments begin with beta-stage robots in 2026-2027 to validate technical integration and operational performance against specific metrics including reliability, serviceability, and ease of integration before scaling to production volumes.
Humanoid plans to offer robots initially through a robot-as-a-service model during the gamma phase, then transition to offering both subscription and capital expenditure purchase options as deployment matures and ROI becomes proven.

Takeaway: This represents a significant shift from pilot projects to production-scale commitments with clear timelines and validation frameworks. The fact that a €16B+ manufacturer is planning hundreds of units within three years suggests the business case for humanoids in manufacturing is becoming concrete enough for major operational bets.

News

X Square secures $140M for 'Physical Foundation Models'

Snapshot: Chinese robotics startup X Square Robot closed over $140 million in Series A++ funding to develop WALL-A, its vision-language-action model designed to give robots the intelligence to handle real-world tasks without extensive programming.

Breakdown:

The company combines vision-language-action models with world models and large-scale real-robot reinforcement learning, allowing robots to learn through physical interaction rather than just simulation or pre-programmed instructions.
Investors include ByteDance, HongShan, Alibaba Group, and Meituan , signaling strong confidence from major Chinese tech players that embodied AI has moved beyond research labs into commercial viability.
X Square demonstrated practical capability with its Quanta X1 robot autonomously completing food delivery tasks in open environments, handling wind, deformed packaging, and visual obstacles without human intervention.

Takeaway: This funding round marks another data point showing that embodied AI for robotics is attracting serious capital and moving toward commercial deployment, particularly in Asia. The shift from teleoperation and simulation to autonomous real-world task completion suggests the timeline for general-purpose robots in logistics and service sectors may be compressing faster than many Western companies anticipate.

News

Autonomous 'Maize Runners' and tractors hit the fields

Autonomous 'Maize Runners' and tractors hit the fields

Image Source: Gemini / There's A Robot For That

Snapshot: Farms are deploying 150-pound autonomous robots and self-driving tractors right now to solve labor shortages and cut costs, with one Ontario farmer saving $45,000 while reducing chemical use by 70%. Demonstrations at Grey Bruce Farmers' Week showed these systems working in real field conditions, not just labs.

Breakdown:

Upside Robotics' autonomous "Maize Runner" cuts nitrogen fertilizer use by up to 70% while improving yields by delivering targeted doses using real-time soil and weather data, though the company admits they're "building a plane while flying it" after robots got stuck in mud and lost in tall corn during early tests.
John Deere's See & Spray technology uses 36 cameras to identify and spray only weeds while traveling at 15 mph, with farmer Mark Ribey reporting $45,000 in savings on his non-GMO soybean crop, though the system requires high sun and clean cameras to function properly.
Fifth-generation autonomous systems now solve about 95% of early deployment problems, but panelists emphasized that startup costs run into tens of thousands of dollars plus ongoing subscription fees, and as one operator noted, "it replaces the seat, not the farmer."

Takeaway: The shift from R&D to commercial deployment is happening faster than expected, with measurable ROI already documented on mid-sized farms. Companies evaluating automation should expect 12-18 month learning curves and recognize these systems augment rather than replace human oversight.

Other Top Robot Stories

TESOLLO launched its DG-5F-S humanoid robotic hand at 60% the cost of its predecessor while maintaining 20 degrees of freedom across five fingers, signaling that proprietary actuator development is driving down component costs faster than many expected in the humanoid supply chain.

WIRobotics advanced from technology demonstration to execution phase at CES 2026, establishing concrete plans for technical collaboration with AI companies on its ALLEX humanoid platform and securing distribution partnerships for its WIM wearable robot across Mexico, the Middle East, and Southeast Asia.

GITAI demonstrated autonomous coordination between two Centaur-style rovers in desert testing that drilled, scooped, and sealed lunar soil samples without human intervention, signaling that multi-robot coordination capabilities for extreme environments are moving from concept to field-tested reality.

LimX unveiled COSA (Cognitive OS of Agents), described as the first physical-world-native agentic operating system that enables its Oli humanoid to perform high-level reasoning and whole-body control, representing a shift from command-based operation to contextual autonomous decision-making in real-time.

🤖 Your robotics thought for today:
PNDbotics ships Adam-U Ultra with 10,000 pre-trained scenarios that deploy new skills in hours, not months—so if the setup barrier just collapsed, why are companies still treating humanoid pilots like multi-year R&D projects instead of quarterly capability upgrades?

Until tomorrow,
Uli

China’s PNDbotics humanoid learns in hours without setup

Great! Check your inbox and click the link to confirm.
Please enter a valid email address.