#CoRL2026 is a wrap. One takeaway from conversations that wasn't in any paper, talk, or poster: robotics researchers have embraced coding agents. Last year, opinions were mixed. This year, nearly everyone agrees they're amplifying researchers.
Robotics is accelerating. π
I'll be at #CoRL2026! I'm on the Atlas team at @BostonDynamics, where I work on VLA research, scaling laws, and model architectures.
If that overlaps with your work, or you're curious about Atlas, I'd love to chat. DMs open.
Teaching humanoids to move heavy objects is way harder than moving light objects. With light objects, a stiff controller is fine, but with heavy objects you have to learn to balance the robot and its payload. This is how you build a robot learning flywheel that will really generalize.
Better motion creates better intelligence.
Atlas can move in more ways, can collect richer data, learn more behaviors, and become more generally capable.
@danwilliamsphil AI learning is probably going to be super different from human or non-human animal learning due to the differences between biological and silicon computers, so it seems like a waste to try to draw inspiration from human or animal learning
Trying to find how and why humans learn is the wrong way to build AI, both are attempts to inject human knowledge into the learning process. It is ok for AI to learn using different mechanisms as they will probably be tremendously more effective.
So much controversy triggered by claims about how humans learn. But the deeper question is why the way we learn works. We need to understand the why to know in what way βhowβ matters. Nature is inspiration for AI, not prescription. The crux of the whole debate is abstraction.
Underneath all learning is representation. The key question is not whether there is a βworld modelβ or whether there is imitation or not. The instrumental question is, what kind of representation does the learning mechanism produce? And then, what are all the leaning mechanisms, natural or not, that can produce good representations?
@ritvik_singh9 Congrats! I was amazed at how good this approach looked running on Atlas MTS. Can't wait to see what the open source community does with it!