Our July Progress Update just dropped:
Orbis 2, a leading World Model built primarily on NATIX data, outperformed several other world-leading models, despite being smaller and trained on le
If you had to teach an autonomous vehicle how to drive somewhere, would you rather show it hours of footage of real-world driving, or an F1 driver doi
Driving in North America is not the same as driving in Europe or Japan.
LLMs taught AI how to understand language. World Models teach AI how to understand physics.
Happy to be a Yuma portco https://t.co/NsYQx8F19W
Our special broadcast, featuring a deep-dive into how we’re building Orbis 2, is coming soon!
An autonomous vehicle trained on expert driver data learns how to make one maneuver.
Turning thousands of hours of real-world driving data into a world model requires more than scale, it requires the right architecture.
You collected data. Congrats. Now, how do you find the moments in your footage that matter?
Collecting data in one city means building autonomy in one city. That's not scalable.
A vehicle approaches an intersection. What happens next?