"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 #Flexionshow more

Read the full technical breakdown here:

If you’ve worked with robots, you know just how challenging many of these tasks are. Incredible work by @FlexionRobotics to bring so many of them together in a single system.

Love the ambition on long-horizon logic.

damn thats impressive easily the longest task ive seen a humanoid robot do and features everything from navigating rough terrain to manipulation and navigation incredibly well done this is starting to feel like robot agi

But can it jack me off?

The combination of reasoning RL for real world robustness looks very promising. Keep pushing the frontier!

Shameless to share our work here if you want to recreate a similar navigation demo like flexion

🔥🔥🔥

@lukas_m_ziegler Impressive

Reminds me of this

Very impressive! Elevator would have left without me if I took that long to press a button.

Incredible 🤯 Future is now coming

incentivised to get investors hope it wasn't remote controlled

stairs 去拿、elevator 上来，这路线有点故意刁难机器人；电梯一忙，snack 先凉半截。

Reflect v1.0's single RL policy should let low-level recovery happen without handing off to a separate controller. 🔄

The path from research to real deployment is shrinking, I wonder how you are planning this next step? Are you already deploying in factories?

👀

hmm interestingg

very cool! i would have left him a parcel with a nice hat or the like 😊

Are those mimic hands ?

How did this work through this long with this much accuracy... will have to read the paper end to end to understand

Impressive to see such complex tasks handled from a single instruction!

It’s interesting to watch both the reasoning as it performs the various steps. But I something that caught my attention was the way it orients on the elevator buttons - how much of that process is determined by the AI and how much by the hand/arm hardware?

didnt know you could change out the heads on those. I wonder why you did that and put the lidar in front. I see the head can move up and down. Thats cool. I wonder if you could build a custom head to rotate 360.