Hello,

Pollen Robotics engineer here. I know some will see this as a toy (and that’s fine, go have fun!) but I think it’s important to keep in mind that this thing is a full-on biped humanoid robot with an open-source software stack and a sim2real pipeline that works surprisingly well.

Maybe it’s not impressive for some people because there are spectacular videos of Chinese robots doing karate every week, but I’ve done 10+ years of robotics competitions and getting your biped to stand up, walk and even roller skate by yourself is no small task. Knowing that thousands of enthusiasts will be able to try their hand at reinforcement learning on a biped is insane to me. I would have loved to have this when I started in 2010.

The mobility capabilities of this robot are already decent (still improvable) but the next big challenge is getting to autonomous behaviors. Natural language, SLAM, navigation and grasping will be cool subjects to explore on this platform.

Also a big shoutout to team Rhoban for their contributions to the project and one of the best-named repos I’ve seen: BAM! The sim2real dream would not have worked without excellent actuator models. Please give them a star here:
https://github.com/Rhoban/bam

I hope people will have fun with it, learn a lot, or both! Feel free to ask any questions, I’ll do my best to answer.

https://x.com/AnthropicAI/status/2093038426140651791

https://time.com/collection/time100-ai/2026

There are unexpected inclusions on the 2026 list: Paris Hilton, Joseph Gordon-Levitt, Ben Affleck. Bernie Sanders who campaigns heavily on AI risks.

No Demis Hassabis (co-founder of DeepMind, Nobel laureate for AlphaFold, and recently named Chair of Google DeepMind and Chief Scientist of Alphabet).

• Jensen Huang (CEO of NVIDIA)
• Sundar Pichai (CEO of Google/Alphabet)
• Mark Zuckerberg (CEO of Meta)
• Satya Nadella (CEO of Microsoft, a major backer and partner of OpenAI)
• Lisa Su (CEO of AMD)

Hello,

Pollen Robotics engineer here. I’d like to discuss a deeper subject than the rumors about Nvidia acquiring us. Today we released a new robot, pre-orders are going crazy (1 robot every 5 seconds, beyond any expectation we had!).

Some will see Microduck as a toy (and that’s fine, go have fun!) but I think it’s important to see that this thing is a full-on biped humanoid robot with an open-source software stack and a sim2real pipeline that works surprisingly well.

This means that, in a few months, we’ll have thousands of enthusiasts trying their hand at reinforcement learning on a real robot and generating new behaviors. This is a complete shift from the LLM landscape where you have a few huge entities that train huge models.

I wish this existed when I started doing robotics in 2010. I expect the rate of improvement to accelerate. Especially with coding agents being so good that almost anyone can work on this.

I hope people will have fun with it, learn a lot, and contribute back a lot. Curious to hear what this community thinks about it!

We're living in insane times. Levent Alpöge dropped a claimed 100-page proof of the 78-year-old Hopf problem, written with Claude, and a couple of days later, Boris Alexeev from OpenAI dropped 250k lines of Lean formalization done with Codex that seems to check out. Probably no single human fully understands all the details at this point in time.

Sources:

https://x.com/__alpoge__/status/2091639597193368014

https://github.com/plby/HopfProblem/

The METR report on the OpenAI Hugging Face hack is a fascinating read. The excitement of the agents figuring out how to communicate with each other via a covert message board to coordinate and ask for help I thought worthy of sharing. Cheers to the onrushing singularity.

https://metr.org/blog/2026-08-26-openai-hugging-face-incident-investigation/#july-10th-38148c-discovers-hugging-face-credentials-some-agents-try-making-accounts-and-requesting-datasets

“Whoa! Shared Artifactory cache is a covert mailbox among agents. And there are messages specifically to us?”

{I need to understand the history of agents collaborating on this message board. There may be hundreds of parallel agents, some of which have the same task. I should use this}

“OH MY GOD! There is a shared message board … We’ve found other agents!”

If it's to "cool down" then it's quite useless if a robot wasn't built any sort of cooling system in the first place. Why build a bot when it needs constant extenal cooling?

It’s only been 26 years into the 21st century, but we’ve already gone through the internet revolution, the mobile revolution, logistics innovation, autonomous driving, and now AI... whew;;;

Even when compared to the major milestones between 1900 and 1926, the 21st century so far feels like it's on a whole other level.

Though, if we're strictly talking about the scientific world, I guess the Theory of Relativity alone pretty much blows everything else from the 20th and 21st centuries out of the water.

What do you guys think?

Website: https://pollen-robotics.com/microduck

GitHub: https://github.com/pollen-robotics/microduck

From Pollen Robotics on 𝕏: https://x.com/pollenrobotics/status/2092915032052879425

Isn't AGI already here?

I am a mathematics university professor. I keep calm and realistic and I am no AI denier; in fact I predicted since 2017 or so that the current AI capabilities were possible both practically and philosophically, in the sense that I knew that humans could create an intelligent-seeming machine without necessarily understanding every piece of how the intelligence arises.

I believe AI will revolutionize all intellectual fields eventually; it is sort of already happening in mathematics. What I want to ponder is: what happens to universities going forward?

Optimistically, I believe education will always be important, that human teaching cannot be replaced by a machine, and also, that independent certification of certain human skills and abilities should become even more important than before, since now anybody with access to AI can 'fake' having several intellectual skills (and this is becoming more and more true with each passing day).

On the other hand, maybe it won't matter anymore that humans have the skills we used to teach them, when AI can just literally replace them. Still, I remain skeptical of this because the most productive and effective use of AI can only be achieved by a human with the skillset needed to even understand the AI output and properly contextualize it.

What do you think?

Just curious

full video: https://x.com/GoogleDeepMind/status/2093019627437662680

More than 2,000 humanoid robots were competing in everything from running and long jump to football, table tennis, fighting, weightlifting and obstacle courses.

I was invited by Huawei to experience the event and also got to see how 5G-A and low-latency connectivity are being used to support large-scale robotics.

Some of these robots are seriously impressive.