Darius Foodeei

(1/N) Days after @RichardSSutton clarified that “brute-force computation is not sufficient,” people are proclaiming that LLMs will inevitably absorb robotics. I think that's dangerously short-sighted.

Fable 5.2 and Mythos 5.2 are a step above Opus 5.5 in almost all the capabilities we evaluated. Testers have found that they communicate more clearly than past Fable and Mythos models. They generate outputs up to 30% faster than Fable 5.1 and Mythos 5.1.

When I visited Boom last week, one of the things that struck me is that they've become the most AI-pilled hardware company I know. With AI coding, every hardware engineer is now a software engineer.

We need France to return to its texhno-capitalist roots More nuclear, more compute, more aerospace, more AI, less bureaucracy Time for 🇫🇷/acc

『造船所で溶接や検査を行う四足歩行ロボット』 足元に電磁永久磁石を採用し、垂直な壁面や曲面状の鋼構造物上の移動が可能。 lnkd.in/gtdePAqF #quadrupedal #robot #PhysicalAI #welding #inspection #shipyard #magnet #EPM #DidenSpider #DidenRobotics

Two months of ablating π0.5 finetunes on a real manufacturing task. The policy now succeeds 98% of the time. We're publishing all results, all the data, and every run, including the ones that went nowhere. What we learned: 1/ 5x more data took us from 63% to 76%. Pure scaling was the weakest lever we found. 2/ Diversity got more out of the same hours. 4h spread across five scenes beat 4h in the eval scene by 30pp. 3/ Quality did even better. One clean, slower hour on top of the 21h: 76% → 90%. That single hour added more than the 17 hours before it. 4/ So we tried quality on its own: 1h of clean data plus 240 rollouts with human interventions. 1.7h in total, 28% → 88%. That's 12 points above the 21h model with a twelfth of the data. 5/ Inference settings add more on top, with no retraining. They took the 21h model from 76% to 93%, and the 21h+1h model from 90% to 98%. Just as useful: what didn't matter. Learning rate, batch size, relative vs. absolute joints, image augmentation. All less important than we expected. Full write-up linked below.

Today we’re unveiling Odyssey-3, a big step forward for foundation world models. It can control robots, power humanoids, drive cars (on the roads of India!), train AIs, pilot drones, and even play video games. We can’t wait to see what intelligent systems it enables.

LLMs may be a dead end to superintelligence.

Starting with the fundamentals Prototype Version 0 AI, Software, Hardware A small team, 9 months Designed and assembled in Paris at @UMA_Robots

"A parcel with snacks has been delivered for Flexion. Retrieve it using the stairs and come up using the elevator. Then unpack it and place the items into the empty drawer on the shelf in the snack area." One instruction. No human operator. Everything that follows is autonomous. Today we're introducing Reflect v1.0, our robotics intelligence platform for long-horizon work. From a single natural-language command, the robot understands the task, navigates a multi-floor building, calls elevators, handles doors, uses tools to unpack a box, and puts the items away. The biggest shift in v1.0 is that we use reinforcement learning across every layer, from low-level control to high-level reasoning. Long-horizon autonomy is unforgiving. The robot must recover on its own when things don't go to plan because in the real world, they never do. Combining reasoning, perception, physical execution and runtime robustness into a single mission-capable system is the foundation required to solve humanoid autonomy. Our team is just getting started. #HumanoidRobots #Flexion

Working in robotics right now is what I imagine working with language models felt like in 2023. Everyone throwing things at the wall to see what sticks Pixel prediction (Cosmos), action prediction (VLA), reward prediction (TD-MPC), and representation prediction (JEPA). Different paths for the same problem The recipe that won in language was self-supervised pretraining at internet scale then light finetune on top. Only representation prediction runs that playbook. It learns from action-free video data so you can pretrain on YouTube and egocentric data then add a control layer. Everything else needs action-labeled data that doesn't scale As an RL maximalist, I used to hate LeCun's cake. Turns out he was right all along which is how I ended up a JEPA truther

The past year has been the happiest period of my life. Genesis is a place of miracles. We set countless ambitious goals. None of them seemed possible with existing technology. One by one, we made them real. The density of breakthroughs coming from such a small team, across so many directions at once, feels almost rare in human history. What you see here is fully autonomous, 1x speed, run on the exact same model.