- Published on
Figure 02 Humanoid Shows Off Rapid, Dexterous Package Handling in New Helix-Powered Demo
- Authors
- Name
- Humanoids daily
- @humanoidsdaily
Figure 02 Humanoid Shows Off Rapid, Dexterous Package Handling in New Helix-Powered Demo
Sunnyvale, CA – Figure CEO Brett Adcock has released compelling new footage of the company's Figure 02 humanoid robot demonstrating significant advancements in logistics tasks, specifically sorting and scanning a variety of packages with impressive speed and dexterity. The demonstration, powered by Figure's Helix AI system, was shared via X (formerly Twitter) in a pointed reply to a skeptical comment.
The exchange began when Adcock posted a video from February showcasing an earlier iteration of Figure 02 sorting packages on a conveyor belt, captioning it, "This was a few months ago...Helix is now showing massive improvements in logistics - can’t wait to show you what’s new." A user named Daniel replied, "This will NEVER WORK with randomly changing packages."
Two hours later, Adcock quoted Daniel's reply and posted the new video with a succinct challenge: "Is this working Dan?"
Is this working Dan?
This will NEVER WORK with randomly changing packages
Advanced Manipulation and Autonomous Learning
The latest video showcases Figure 02 operating at a notably faster pace and with greater finesse. The robot is seen autonomously handling a stream of diverse packages – varying in size, color, material, and deformability, appearing to be real postal items. With fluid movements, Figure 02 picks up each package, orients it to find the barcode (including turning packages over), and in the case of deformable items like polybags, it has learned to flatten them against the scanning surface to ensure a successful scan – a behavior Adcock noted is akin to how a human would perform the task.
In a follow-up post, Adcock confirmed, "This is fully autonomous driven by Helix. The policy is flipping packages to orientate the barcode down and has learned to flatten packages for the scanner (like a human would). More coming soon."
This demonstration highlights the capabilities of Figure's Helix AI model, which, as we previously reported, is a Vision-Language-Action (VLA) model designed for generalist control and learned behaviors. The ability to handle unseen, deformable objects and adapt techniques like flattening packages underscores Helix's learning capabilities and its potential for real-world logistics applications. Indeed, such advancements are likely what has prompted logistics giants like UPS to reportedly explore deploying Figure AI's humanoids within their network, as covered by us in April.
Industry Reactions and Performance
The video quickly drew attention, with humanoid robotics expert Scott Walter tweeting, "'But it's not as fast as a human,' you say? Yes, some workers are faster. But compared to the average it is faster because it has a higher success rate. F.02 is already more efficient than the average. And it's the worst it will ever be." Walter also added a teaser, "And F.03 is still in the Bear Cave," referencing Figure's next-generation humanoid.
"But it's not as fast as a human," you say? Yes, some workers are faster. But compared to the average it is faster because it has a higher success rate. F.02 is already more efficient than the average. And it's the worst it will ever be. And F.03 is still in the Bear Cave
Is this working Dan?
Figure has been making rapid strides. The company recently reorganized to consolidate its AI efforts under the Helix banner, aiming to accelerate robot learning. This new package handling demonstration is one of the most concrete examples to date of Helix's application in a dynamic, commercially relevant task.
Towards F.03 and Broader Deployment
The progress with Figure 02 and Helix AI comes as the company is also developing its next-generation F.03 humanoid. As we covered in May, the F.03 model is already walking and is being designed for affordability and high-volume manufacturability.
While the latest video is a significant step, questions regarding long-term reliability, adaptability to an even wider range of package variations, and overall integration into complex warehouse workflows will be areas to watch. However, this demonstration provides a strong indication of Figure's trajectory and the increasing sophistication of its humanoid robots for practical applications. The company's focus on end-to-end learning with models like Helix appears to be yielding tangible results in complex manipulation tasks.