Saw Neal Stephenson in Harvard Square last night for a reading of D - sequel to Polostan.
I brought this well loved book for signing. Bought it back in 1992 in high school at I think the Mr. Paperback in Waterville, Maine.
Maybe helped steer my career in subtle ways.
chrisbentzel.com/posts/micro…
Microstates - can a US State consist of only one person? Can a political party use small states to gain a permanent stranglehold on electoral base?
Pretty different from my normal topics, but fun one to explore and write.
New post - a review of Inference Engineering by @philipkiely. chrisbentzel.com/posts/infer…
Biked to and from office on Friday - ~60 miles total. Not an everyday plan at 4 hours of commute time, but really enjoyed it.
Chris Bentzel retweeted
"Now that the Atlas enterprise platform is getting to work, the research version gets one last run in the sun."
Farewell to the OG electric Atlas, a true engineering marvel!
Chris Bentzel retweeted
Google DeepMind 🤝 @BostonDynamics
Our new research partnership will bring together our advancements in Gemini Robotics’s foundational capabilities to their new Atlas® humanoids. 🦾
Find out more → goo.gle/49paguA
Chris Bentzel retweeted
I don't get the "ChatGPT will kill Google" takes. They have the training data, compute resources, and know how. Great distribution too. Main risk likely fear of fair use implications more than cannibalizing business. Knowledge Panels and Assistant already point this direction.
Chris Bentzel retweeted
VLAs have become the fastest-growing subfield in robot learning. So where are we now?
After reviewing ICLR 2026 submissions and conversations at CoRL, I wrote an overview of the current state of VLA research with some personal takes:
is.gd/1pqw9w
Chris Bentzel retweeted
Working side by side with robots? Safety is key. 🤖
Matt Powers from @BostonDynamics shares his quick take on innovation, policy, and safety best practices.
Full panel recording coming soon. #NVIDIAGTC
Chris Bentzel retweeted
Contact-rich manipulation is the correct next frontier for robotics and this is one of the cooler examples around. It's so clear that in the next couple years robots will be dramatically more capable of the kind of challenging, contact-rich tasks that they're currently bad at
See Spot perform dynamic whole-body manipulation. Using a combination of reinforcement learning (RL) and sampling-based control, the robot is able to autonomously drag, roll, and stack tires weighing 15 kg (33 lb), well above its maximum arm lift capacity.
Learn more about coordinating locomotion and manipulation processes: rai-inst.com/resources/blog/…
Advancements on benchmarks like these will be a key step to physical robotic agents.
Chris Bentzel retweeted
The first large language health model that predicts >1,000 diseases and the time they will occur in a person was just published @Nature, by @MoritzGerstung @artem_shmatko and colleagues nature.com/articles/s41586-0…
This, along with other recent groundbreaking reports, represents the dawn of a new era of primary prevention in medicine
Chris Bentzel retweeted
🚀 Boston Dynamics is growing!
We’re hiring Research Scientists & Engineers to push the limits of humanoid robotics 🤖🧠.
If you’re into reinforcement learning, large-scale physics simulation and dexterous manipulation - we’d love to talk.
1/3
Excited to show off our end-to-end learning for whole body manipulation on Atlas, done in collaboration with TRI!
youtu.be/HYwekersccY?si=1sT5…
Chris Bentzel retweeted
I wrote a fun little article about all the ways to dodge the need for real-world robot data. I think it has a cute title.
sergeylevine.substack.com/p/…




















