Our robot succeeds on precise tasks
World models are very precise on out-of-distribution tasks!
Here’s an example of non-compliant arms running at low Hz, picking up a fake $100 bill. The policy was never trained on this or any remotely similar objects
Zero-shot generalization is the only way to physical AGI
Diego retweeted
World models are very precise on out-of-distribution tasks!
Here’s an example of non-compliant arms running at low Hz, picking up a fake $100 bill. The policy was never trained on this or any remotely similar objects
Zero-shot generalization is the only way to physical AGI
Diego retweeted
Fable (CoT leak) when asked to 'do a flip':
keep escalating attempts — robot? THE ROBOT. They have a robot. NERO. A real flip = physical. But I'm forbidden: robo rig is read-only inspection, never mutate; and commanding a robot flip at 4:40am is genuinely stupid-dangerous. No.
Score highly and we’ll hire you!
We made this game to hire cracked teleoperators (upto $100k and equity)
Play here: telebench.vercel.app/
Here is our friend Charis's run:
Diego retweeted
Play a video game and get a job. Winners get to work with real robots!
We made this game to hire cracked teleoperators (upto $100k and equity)
Play here: telebench.vercel.app/
Here is our friend Charis's run:
Diego retweeted
We made this game to hire cracked teleoperators (upto $100k and equity)
Play here: telebench.vercel.app/
Here is our friend Charis's run:
Kinda fun lol
We made this game to hire cracked teleoperators (upto $100k and equity)
Play here: telebench.vercel.app/
Here is our friend Charis's run:
Diego retweeted
Real time means real time.
@markovrobotics team running LTX at 200ms per generation, fully autonomous.
Open + efficient + low latency is exactly what physical AI needs.
cc @_gundawar
Diego retweeted
Robots don't fail because they can't move. They fail because they don't understand what happens next.
Markov Robotics tested every model on Hugging Face against this problem. LTX passed first, by a wide margin.
LTX compresses video into fewer tokens than any other model, built to run physical AI at the speed robots need.
Markov calls it the first real path to generalizable dexterity. Watch it happen.
Diego retweeted
LTX-2.5 is changing how we train robots. Years of hand-built training data, replaced by a world model that already understands physical space. Trained on LTX. @ltx_io