Will World Models Eat Physical AI?

For years, robotics teams worked around a fundamental constraint: you could only teach a robot what you could directly demonstrate or simulate. World Models change that paradigm entirely.

We're bringing together the researchers and engineers closest to this work, including Jason Ma from Dyna Robotics and Chris Paxton from Agility Robotics, to dive into the conversation we have all been itching to have around World Models.

What we're getting into:

• What robotics teams did before world models, what changes now, and which architectural paradigms are showing the most promise
• Does a world model need to understand everything, or is specialisation the real differentiator? And how far does compositional generalisation actually get you?
• How do world models change the way Physical AI teams should think about real-world data, and what emerging data recipes are producing production-capable systems?
• Do VLAs still have a place alongside world models, what is the interaction with reasoning, and what are the concrete challenges the field hasn't solved yet?

Register to attend live or receive the recording.