upvote
The theory is this will ultimately give robots foresight in the physical world, like we can imagine an action before taking it.
reply
Sounds really bad for latency. Seems better to reason in the abstract or produce a mathematical model.
reply
Once a model gets good enough, etch it directly onto silicon and get 50,000 tokens per second.
reply
Aye basically use it to make synthetic training data.
reply
Training in a physics engine like MuJoCo with deterministic control seems more reliable.
reply
That’s.. actually insane. Just imagining a Claude operating a robot arm with a tool call that leveraged something like this to “imagine” scenarios before acting.

My god we are so cooked.

reply
We've seen Waymo training and potentially other use cases: https://waymo.com/blog/2026/02/the-waymo-world-model-a-new-f...
reply
it is for building world models for robots to interact with without killing entire the state of the run (or destroying the robot)
reply
deleted
reply
Build realistic worlds -> Unleash robots in them to have more realistic, synthetic training environments. You know, just like the current world us humans live in. Turtles all the way down and all.
reply
> love of the craft

The feeling of AI recently is Lovecraftian indeed.

reply