Zohreh Dannenhauer ✈️ CORL (@zoedan4) header
Zohreh Dannenhauer ✈️ CORL (@zoedan4) profile picture

AI Engineer

San Francisco, CA
Joined January 2020
Photo by Zohreh Dannenhauer ✈️ CORL (@zoedan4)Photo by Zohreh Dannenhauer ✈️ CORL (@zoedan4)Photo by Zohreh Dannenhauer ✈️ CORL (@zoedan4)Photo by Zohreh Dannenhauer ✈️ CORL (@zoedan4)Photo by Zohreh Dannenhauer ✈️ CORL (@zoedan4)
My favorite slide from the talks so far from Goldberg’ talk #CoRL2026
1
5
64
The debate at CORL: “current recipe of behavior cloning pre-training + in-domain fine-tuning will lead to commercially viable generalist robots." #CoRL2026
1
4
120
Zohreh Dannenhauer ✈️ CORL retweeted
Make your Hermes agent skeptical, so it can tell real from marketing, propaganda, and slop. paranoy for @NousResearch Hermes: your agent weighs the evidence for and against before it believes anything. Demo: loom.com/share/2fa536fdb3094… paranoy.org/hermes
3
8
1
22
409
Zohreh Dannenhauer ✈️ CORL retweeted
synaptic pruning: your brain removes what you don't keep on repeating. when you were a child, joy was easy. ice creams, puddles, first snow. you didn't need to try; you just had to feel them. however, as you grew older, joy became mundane. magic became naïve. lightness became irresponsible. as you may have realized, it's not about life; it’s about biology. your brain removes any connections you don't keep using. no malice, just efficiency. so what do you think you have stopped doing? being spontaneous. laughing with no purpose. believing that good things will come. the good news is that pruning isn't permanent. neuroplasticity is there to help you recreate all of this, not with force of will but with repetition. laugh with no rhyme or reason. do something without a goal. assume that good things will come to you. it might seem fake at first but eventually it will start working.
10
202
36
1,080
174,474
Zohreh Dannenhauer ✈️ CORL retweeted
What will be the “RLHF” moment for robotics? What will it take to get robotics to where LLMs are today and beyond? New blog post with @chelseabfinn sharing some thoughts on the state of RL for frontier robotics models and what's missing 👇 Blog: pd-perry.github.io/posts/pos…
22
87
14
523
130,624
Had so much fun playing with these @trossenrobotics robots! #sfevents @
1
36
What US-based humanoid robot is available to buy with an open SDK? #ChineseRobotBan
1
29
Zohreh Dannenhauer ✈️ CORL retweeted
196
1,191
237
5,098
1,240,726
We have the same experience using VLAs - the solution is not always more data! They don’t generalize. You change the scene slightly and they are useless
We’re going all in on World Models. Today we’re launching the 1X World Model Lab. The bet is simple: You can’t fine-tune your way to AGI. And you definitely can’t fine-tune your way to robots that can operate in the physical world. General-purpose humanoids need models that understand space, motion, objects, causality, affordances, physics, and action before they ever see a specific task. The frontier is not better VLA wrappers. The frontier is embodied world models. The 1X World Model Lab will focus on large-scale embodied world model pretraining: building the most generalizable foundation model for humanoid robots from the ground up. The next frontier in AI requires scaling: web-scale media + egocentric human videos + sim + dexterous remote operated robot data + on-policy NEO data → real-world deployment for robot data collection and RL → abundance of data → physical AI The robot collects data. The model gets better. The robot gets better. Repeat. To lead this, we brought in one of the best for the mission: @_sam_sinha_ , as Head of World Models. Sam was a founding research scientist at Luma AI and has been at the frontier of scaling multimodal generative video models his whole career. If you’re the best in the world at large-scale pretraining, video models, robotics, RL, infra, or data — and you want your models to move atoms, not just pixels — join us. Send background + evidence of exceptional ability to: [email protected] We’re building the model that makes autonomous labor real.
73
“AI bots are scheming together”—really though? It’s a sexy headline, but these bots don’t have goals or self-preservation instincts. #Moltbot
58
Zohreh Dannenhauer ✈️ CORL retweeted
On December 8, the Perseverance rover safely trundled across the surface of Mars. This was the first AI-planned drive on another planet. And it was planned by Claude.
290
836
366
7,820
1,618,806
Zohreh Dannenhauer ✈️ CORL retweeted
JACK DORSEY'S NEW APP IS A PROTESTER'S SECRET WEAPON Bitchat doesn’t need Wi-Fi, mobile data, or even a working government. Messages bounce phone-to-phone using Bluetooth; no towers, no servers, no kill switch. It’s ideal for places like Iran, where regimes love blackouts and protesters need a way to talk when the internet “suddenly disappears.” No logins. No SIM cards. No surveillance. Source: @CatiaKyen
1,465
4,095
591
22,896
1,412,977
Zohreh Dannenhauer ✈️ CORL retweeted
BREAKING 🔴🔴 Iran’s government said the nationwide internet blackout will remain in effect at least until the Iranian New Year in late March. CC: @realDonaldTrump @marcorubio
481
1,556
241
5,036
390,022
Zohreh Dannenhauer ✈️ CORL retweeted
Iran has been in a complete connectivity blackout for the past few days (no mobile phone, internet, landline, etc) as the regime has been murdering citizens in the thousands. Hey @Apple: please make satellite messaging available to Iranian iphone holders so that they can communicate with the outside world!
100+ hours: Iran has no internet, calls, or messaging. Millions hold iPhones with satellite capability, but it’s disabled there. @apple @tim_cook please enable satellite messaging for Iran, even temporarily. People deserve a way to say “I’m alive.”
2
39
1
183
19,022
“Simulating a world is not the same as understanding it.” Models like Genie predict pixels. Great for interactive video, but prone to physical hallucinations. Dreamer and JEPA predict latent states. They are built for agents to imagine thousands of futures to solve a task. An intelligent agent doesn’t need to know every detail in a scene to know how to act. High-definition generation is a distraction; intelligence lies in abstracting the world, ignoring what’s irrelevant, and not recreating it pixel by pixel. #AGI #AI
33