vision-language-action
MIT
+7
Aug 12, 2026
•
6 min read
Teaching Robots to Think Ahead Instead of Freezing to Think
Most AI robots move like a buffering video: watch, freeze, think, act, repeat. A new MIT method called VLASH has them predict where they will be by the time they finish thinking, so they plan the next move while finishing the last one, smoothly enough to hold a real ping-pong rally with a human.
NVIDIA
+10
Jul 15, 2026
•
5 min read
One Free Download Now Drives 20 Different Robots
Ant Group open-sourced LingBot-VLA 2.0, a free model its makers say drives 20 different robot bodies on a single gaming graphics card. A day later NVIDIA and Hugging Face answered. The race to become the Android of robots is on, and nobody is charging.