4 min read

🤖 LeCun Pushes Robot World Model

Plus: Robot Warehouse Brain, Tracks Smell Without Sensor, McDonald’s Deploys Humanoid

Good Morning, Roboticists!

Perception is becoming execution—they learn, adapt, and coordinate in the real world.


LeCun Pushes Robot World Model

TL;DR: Turing Award winner Yann LeCun and researchers from Mila, NYU, and Samsung introduced LeWorldModel, a compact JEPA-based architecture designed to train robots directly from raw visual input. With just 15M parameters, it runs on a single GPU, delivers fast planning, and achieves strong results on robotic tasks, pointing toward practical, perception-driven robot intelligence beyond LLM-based approaches.


Otto Builds Robot Warehouse Brain

TL;DR: Otto Group is deploying AI-driven coordination across its warehouses using Nvidia Omniverse and a custom Coordinated Autonomy Layer (CAL) to manage fleets of robots. By creating a digital twin with Boston Dynamics’ Spot and simulating layouts, the company improved robot efficiency and reduced stoppages, moving toward a system where multiple robots operate together under a unified control layer with human oversight.


Robot Tracks Smell Without Sensor

TL;DR: Researchers from the National Institute of Informatics, Science Tokyo, and Tohoku University built a bio-inspired robot that can track odor sources even after losing one of its two sensors. Modeled on silkworm moth behavior, the system adapts to incomplete input and maintained nearly identical performance in indoor and outdoor tests, unlike traditional robots that fail when sensor balance is disrupted.


McDonald’s Deploys Humanoid Robots

TL;DR: A McDonald’s outlet in Shanghai deployed humanoid robots developed by Keenon Robotics to greet customers, assist with ordering, and deliver meals. The robots, dressed in McDonald’s uniforms, interacted with diners and performed service tasks in a pilot setup, with multiple units operating behind the counter and on the restaurant floor.


Unitree G1 Learns Tennis

TL;DR: A team from Tsinghua University, Peking University, and Galbot developed the LATENT framework, enabling the Unitree G1 humanoid robot to perform multi-shot tennis rallies with humans. The system learns from imperfect motion fragments and uses a constrained control approach to handle fast incoming balls over 15 m/s, achieving stable, real-world performance after simulation training.


Hyundai Builds Humanoid Welders

TL;DR: Hyundai partnered with U.S. robotics firm Persona AI to develop humanoid robots for welding tasks in shipyards. Under a joint agreement with its shipbuilding and robotics units, Persona will design a bipedal platform while Hyundai handles integration and training systems. A prototype is expected by late 2026, with testing and phased deployment across shipyards planned from 2027.


Robot Races Reach Global Events

TL;DR: Humanoid robots participated in athletic competitions at the World Humanoid Robot Games in Beijing, including marathon-style races alongside human runners. The development is extending to the U.S., where the Professional Robotics League (ProRL) will host its first event in Boston, featuring humanoid robots competing in a public 50-meter sprint during marathon weekend.


Drone Swarms Reach Milestone

TL;DR: Palladyne AI and Draganfly completed a key milestone by integrating SwarmOS software with mission-ready drones and validating it through flight simulation. The system enables decentralized drone swarms where each unit can perceive, decide, and collaborate in real time without centralized control, adapting to changing conditions like communication loss or asset failure.


READ MORE

Let the Future Come to Your Inbox

Stay ahead without drowning in information. We turn the most important signals across AI, tech, marketing, and future products into 5-minute reads you can actually finish.


TOGETHER WITH US

AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, reaching over 2 million leaders across the global innovation ecosystem, from OpenAI, Anthropic, Google, and Microsoft to top AI labs, VCs, and fast-growing startups.

We've helped promote over 500 Tech Brands. Will yours be the next?

Email our co-founder Mark directly at mark@aisecret.us if the button fails.