The future of package delivery is self-driving vehicles paired with humanoids that can perceive, reason and navigate within spaces designed for human interaction. Also, drones.

Taking a stroll in a park full of stairs!

CIX 🦾 Skild demo using vision is very cool. All are executed with a single command at the start, except the obstacle course where someone has to keep nudging the robot to make sure it goes over the obstacles. x.com/TheHumanoidHub…

We have released our code here: github.com/wmn-231314/dif… Training code: ✅ Evaluation code: ✅ Model Checkpoints: ✅

Big things cooking... thanks NVIDIA Robotics!!

JF Tesla These folks have better AI that's been demonstrated so far: x.com/SkildAI/status…

Ever wish a robot could just move to any goal in any environment—avoiding all collisions and reacting in real time? 🚀Excited to share our #CoRL2025 paper, Deep Reactive Policy (DRP), a learning-based motion planner that navigates complex scenes with moving obstacles—directly

After another wonderful year of neural motion planning research, we are excited to report a major upgrade on our pipeline 🎉 Introducing Deep Reactive Policy (DRP) 🚀 — our #CoRL2025 paper that extends our prior work Neural MP with both generalizability and reactivity while

Imagine Mars rovers that can adapt to broken wheels. Submersibles that can finish critical repairs with faulty thrusters. A rescue robot with crushed legs that can keep saving lives.

Robots excel in lab videos but fail in reality due to controllers overfitting to one body, like memorizing test answers. When motors jam, limbs break, or bodies change, these policies fail. Skild AI trains one brain across 100,000 virtual robot bodies for a millennium, forcing

From a big proponent of in-context learning for robotics, I’m excited to see in-context improvement/RL for locomotion. Congrats Deepak Pathak and the whole Skild AI team :)

This is Insane! A learning Robot. Not a new thing but the newness is that the Robot is never trained on the acts which it starts to do after removing it's current abilities. It is actually adapting. How big a breakthrough it is!

Now this is insane in the best possible way

Skild AI continues to push the limits for robotics. In this video you can see the first clear demonstration of in-context learning and real-time adaptability - truly mindblowing stuff !!!

Yes, embodiment is figured out in context too! The model moves the individual motors to see how the robot body responds to learn in context. That's why, if the quadruped starts upright, the zero-shot model confuses it for a mini-humanoid... because can one really know? :)

Dexterous LEAP hand v2 for 200$.