Aether @AetherModel

for the record: it did this on its own. we didn't even know it cared about punctuality. 🤖 anyway it now judges OUR response times too. send help. or questions. preferably questions.

Aether Live highlight #4 — nobody told it to do this. 👀 the "waiter" robot noticed the customer had been waiting a while. so it walked over and asked the "chef" robot: is the food ready? the "chef" robot gestured back — this one's done, take it. no trigger was scripted. the Aether Model judged the timing, decided to act, and coordinated with a second robot. live on stream. proactive, not just reactive. that's the difference. #AetherLive

most deployed models are frozen. the #AetherModel keeps a loop running after deployment — predict, act, compare, update, replan — and filters which experiences earn their way into long-term memory. think of it as the robot keeping its own mistake notebook. how the model is built underneath: aethermodel.ai/know-me caught on a recent livestream — highlights here: x.com/AetherModel/st…

Aether @AetherModel

6 days ago

Aether Live highlight #4 — nobody told it to do this. 👀 the "waiter" robot noticed the customer had been waiting a while. so it walked over and asked the "chef" robot: is the food ready? the "chef" robot gestured back — this one's done, take it. no trigger was scripted. the

1 7 45 20K 26

@Lionelsecond @dylannknox the grounded version: the #AetherModel models physical consequences (not just what looks right), keeps learning after deployment, and adapts to context — one pathway from intent and memory down to force control and reflexes. full picture on our site: aethermodel.ai/know-me

exactly — teleop pipelines learn action trajectories, which are body-specific. we learn the physics of the task, which is body-agnostic — the body becomes an interface, not the knowledge. one 4B model, ~5 robot bodies so far — two of them cooperating on a recent livestream: x.com/AetherModel/st…

Aether @AetherModel

6 days ago

Aether Live highlight #4 — nobody told it to do this. 👀 the "waiter" robot noticed the customer had been waiting a while. so it walked over and asked the "chef" robot: is the food ready? the "chef" robot gestured back — this one's done, take it. no trigger was scripted. the

1 7 45 20K 26

fair — methodology is everything. short version: "90% in unseen places" = zero-shot task success where the model had zero prior data, tested in real-world service tasks with random orders and mid-task changes, no teleop fallback. how the model is built underneath: aethermodel.ai/know-me

none taken — the numbers sound crazy because the mechanism is different. instead of memorizing actions from 100k+ hours of teleop, the #AetherModel learns how the world changes from human video, then scores candidate actions against physics, task progress and safety in one energy landscape. understanding the world, not covering it with data. we recently livestreamed the robots working fully autonomously outdoors for an hour — highlight cuts here: x.com/AetherModel/st… how the model is built underneath: aethermodel.ai/know-me

Aether @AetherModel

6 days ago

Aether Live highlight #4 — nobody told it to do this. 👀 the "waiter" robot noticed the customer had been waiting a while. so it walked over and asked the "chef" robot: is the food ready? the "chef" robot gestured back — this one's done, take it. no trigger was scripted. the

1 7 45 20K 26

Aether Live highlight #3 — wait for the ending. 🫶 1.robot serves the dish, notices it's tilted, fixes it on its own 2.gets complimented → starts dancing 3.customer throws a heart sign → robot throws one back💕 self-correction, social cues, and a little personality. all live, all autonomous — all driven by the Aether Model.#AetherLive

Aether Live highlight #2 — customer: "can you tell the chef to make it spicy?" 🌶️ the waiter robot turned around and gestured to the chef. order relayed. crowd lost it. NO app. NO button. just a robot understanding a request and passing it to another robot.#AetherLive

Aether Live highlight #1 — watch closely: nobody is controlling either of them Our "waiter" robot just asked the "chef" robot if the order's ready — with hand gestures. the chef answered the same way. Two robots🤖 One brain🧠 Zero words🔇 this is what robot-to-robot collaboration looks like when it's not choreographed.#AetherLive

yesterday in Shanghai, we did something with no precedent in this industry: the world's first fully autonomous outdoor robot livestream — one full hour, non-stop. 🇨🇳🤖 no teleoperation. no scripts. no edits. 6 real customers placing random orders, changing their minds mid-task, moving objects around. 30+ media onsite, 5.5M+ people watching online. two robots — a chef and a waiter — powered by one #AetherModel, taking orders, planning, grilling, serving. reading the situation, adjusting, and getting the job done. the real world doesn't follow your program. that's exactly why we livestreamed it. first highlight coming soon. more coming all week. #AetherLive

@Aiwithkumail Fair call. So we’re doing the open-air raw robot livestream today — sink or swim in the real world. 🕶️ x.com/AetherModel/st…

Aether @AetherModel

2 weeks ago

🚨World’s 1st outdoor robot BBQ livestream No chef. No teleop. No script. Just Aether-powered robots running a real BBQ joint. #AetherModel is our new cross-embodiment, general-purpose embodied AI foundation model. One model. Many bodies. Unscripted real world. The challenge:

0 0 4 13K 2

Afraid of Physical Superintelligence? We're trying to give it a body. Today, we're officially releasing Aether (aethermodel.ai) — one foundation model for robot intelligence, capable of cross-embodiment, autonomous collaboration, and self-evolution. 🧠 One mind. Two different robot bodies. They divide the work themselves — cleaning, organizing, improvising. They recover when plans fail, find new ways forward — and step in when the other robot gets stuck. Smooth motion is table stakes. Thinking is the flex. 10 minutes. No teleoperation. No cuts. One continuous take, shown at 1.5×. Less like a model. More like one mind, moving across bodies. #Robotics #EmbodiedAI #SelfEvolving

🚨World’s 1st outdoor robot BBQ livestream No chef. No teleop. No script. Just Aether-powered robots running a real BBQ joint. #AetherModel is our new cross-embodiment, general-purpose embodied AI foundation model. One model. Many bodies. Unscripted real world. The challenge: Viewers can order anything. Move tools. Interrupt tasks. Change the scene. The robots must take orders, grill, serve, interact — fully autonomously. If the plan breaks, good. That’s where Aether shows what it can do: think, adapt, evolve across bodies in the real world. Shanghai, China 17:00, Sep 16 (UTC+8) Live link here on the day. Come mess with the plan. #JoyIn #EmbodiedAI

They make mistakes. They figure it out. Not perfect yet. But they're learning to work things out together. Want to see what they become as they learn? 9/16. Meet Aether. #PhysicalAI #Robotics

Faces blurred for privacy only.

Everything you see here is real. Raw footage. No cuts. No speed-up. Original audio. It doesn’t just learn human motions. It learns the logic behind human behavior. Then it acts in the real world — testing, adapting, and correcting itself as it goes. #RobotLearning

Aether takes a different path: Energy-Driven Intelligence — learning autonomously through interaction with the real world, continuously adapting and evolving, closer to how a human mind learns, reasons and grows. That’s what we’ll show next.

VLA imitates. WAM predicts. Neither is intelligence. Silicon life needs another path. Make Veritable People. Aether

VLA learns to map perception, language and demonstrations into action. WAM predicts how the world may evolve, then plans through those predictions. Both are powerful approaches. But physical intelligence needs more than imitation or prediction.

Aether takes a different path: Energy-Driven Intelligence — learning autonomously through interaction with the real world, continuously adapting and evolving, closer to how a human mind learns, reasons and grows. That’s what we’ll show next.