← Home
TWIML · September 1, 2026 · 1h 6m

World Models and the Future of Spatial AI with Justin Johnson - #775

Justin Johnson
Show Notes

Justin Johnson, co-founder of World Labs, explores the emerging field of spatial AI and world models—systems that can understand, generate, and simulate 3D environments. He discusses the divide between explicit 3D representations and generative approaches, and why there is still no consensus on how to build these models. The conversation covers World Labs’ Marble system, which creates navigable 3D worlds from images, the open challenges of evaluating world models, and the role of simulation, planning, and action. Johnson also shares his vision for unified models that could power interactive virtual environments and agents that operate in the physical world.

This summary was generated from show notes and public descriptions, not from a full transcript review. Details may contain inaccuracies.

Curious

•
World models: no established recipe
Building world models remains an open research problem; there is no settled approach, with competing paradigms like explicit 3D representations and generative models.
ClaudeChatGPTGemini

Highlights

•
Beyond language: spatial AI as the next frontier
Many researchers now see capabilities beyond language—particularly spatial understanding and world modeling—as an important frontier for AI.
ClaudeChatGPTGemini
•
Marble: generating navigable 3D worlds
World Labs’ Marble system can generate navigable 3D environments from a single image or other inputs.
ClaudeChatGPTGemini
•
Evaluating world models is an open challenge
There is no widely accepted framework for evaluating the quality and usefulness of world models.
ClaudeChatGPTGemini
•
Unified world models for agents and robots
Justin Johnson envisions a future where world models unify perception, planning, and action, enabling both virtual agents and physical robots.
ClaudeChatGPTGemini
Was this useful?