James Wigglesworth

CEO - Amplibotics AI
Making a future where robots are actually useful.

[Igor Kulakov (MicroFactory) @ihorbeaver]

Benefits of using a reliable VLA model on example of our customer’s task.

Instead of building special tooling, we use the model to separate the top magnet from the stack directly in the grippers.

In the demo, the model moves the stack up, one magnet at a time, visually finding the right height. It also corrects the slight randomness in the magnets’ positions so the top magnet ends up in the same place, where it’s easy to pick it off.

When the stack is almost empty, the model sees it as the right gripper covers the stack, and later it will go get a new batch, but for now, in this test, it just throws it away.

No coding, except saving points
1x speed, autonomous

[James Wigglesworth @wiggleswor27016]

@KyleVedder Inspired by this post a month ago and came back to say we've now implemented this!
Policy in the loop validation, DAgger style interventions.
Post-training data from real robots doing real tasks either on our 'Robot Arm Data Farm' or deployed in the real world.
Shoot me a DM.

[James Wigglesworth @wiggleswor27016]

@sean_pixel @KyleVedder This is exactly how we do it!
Centralized robots, de-centralized operators.
Things work better when robots stay near engineers, near-shore operators create scalability.
There is a range of about a half-hemisphere diameter before latency becomes an issue.
https://t.co/WUE9kYgwv7

[James Wigglesworth @wiggleswor27016]

What's better: 4 fingers or 5 fingers?
5 is closer to humans and might transfer human video data better.
But 4 fingers is the simpler mechanical solution without really losing that much capability

[RoboHub🤖 @XRoboHub]

UCLA RoMeLa just open-sourced a dexterous hand that hits the rare combo: tactile sensing, humanlike form factor, and a roughly $3K build cost. 🖐️

MIDAS packs 16 DoF, 13 active + 3 passive DoF architecture, 283 3-axis tactile taxels, open-source hardware and software, a MuJoCo contact model, and a vision-based teleoperation pipeline.

The real value is not just price; it is a repairable, tactile-rich hand built for manipulation, teleoperation, and robot learning work that usually costs far more to prototype.

If dexterous manipulation is going to leave locked lab stacks, this is the kind of hardware more researchers need on the bench.

[Amplibotics AI @amplibotics_ai]

Physical AI needs physical data.

Amplibotics operates a fleet of tele-operated robots generating manipulation datasets for frontier robotics teams.

Real robots. Real humans. Real work.

Need data for your next VLA, policy, or foundation model? Let's chat.

#PhysicalAI #VLA

[jonathanhawkins @jonathanhawkins]

For this hackathon I taught a robot to beat you in Jenga.

It sims every possible move, picks the block that's the best move, then an RL policy pushes it out.

Built in MuJoCo & @threejs, running on a real @AgilexRobotics arm.

[CyberRobo @CyberRobooo]

Approaching Human-Level Dexterity

Beijing-based DeepCybo’s Prime humanoid robot has achieved smooth and precise tool manipulation for household tasks, such as chopping vegetables, cutting cake, stirring eggs, and peeling cucumbers.

It is driven by their Z-WM (World Model) and executed by the Wuji Hand dexterous hand.

Interestingly, DeepCybo mentions that they train the World Model using human data.

The effectiveness and reliability of humanoid robots ultimately come from the full-stack synergy of data, models, and hardware (including dexterous hands).

[The Linux Foundation @linuxfoundation]

Rohit John Varghese, Director of Systems Engineering and Product at Contoro Robotics, is live on the #OSSummit keynote stage, discussing how MCP bridges the gap between digital AI reasoning and physical robotic execution.

[Vaishabh Jalmi @VaishabhJ]

Open-source robotics is getting insane 🤯🦾 This is the SSG-48 Adaptive Electric Gripper — a fully open-source robotic gripper with force control, ROS2 support, Python API, and 3D printable parts. Perfect for robotics projects, AI robots, automation, and makers building futuristic hardware at home. 🚀 SSG48 adaptive electric gripper Project by Source robotics 👇 GitHub Link -SSG-48 Adaptive Electric Gripper GitHub https://t.co/U0NfIFSZic #robotics #opensource #engineering #robotarm #robotgripper automation airobot 3dprinting arduino raspberrypi mechatronics roboticsengineering opensourcehardware maker tech futuretech embedded electronics ros2 pythonprogramming

[Shintaro Inoue / 井上信多郎 @shin0805__]

A wheeled-legged robot that can float, move in 3D, and use its legs as arms.

Our paper on the new robot "WiXus" has been accepted to #ICRA2026!

By anchoring wires to the environment, Wixus expands wheeled-legged robots beyond locomotion-toward manipulation and tool use.

[BitRobot 🦾 @BitRobotNetwork]

In robotics, data volume alone isn’t enough. Quality, geographic diversity, and dynamic real-world environments matter.

RoboCap is our hardware platform designed to capture the richest egocentric datasets for embodied AI.

Explore the new site: https://t.co/B0ejMdf3j9

[ViralRush ⚡ @ViralRushX]

A humanoid robot learning hand eye coordination through juggling and object catching.

[James Wigglesworth @wiggleswor27016]

Amplibotics AI - who we are and what we're doing

[Ilir Aliu @IlirAliu_]

15–30 minutes of real-world robot data.

That's now enough to go from

sim-to-real failure to working robot.

Let’s see…

You train a robot in simulation. You deploy it in the real world. It fails. The physics don't match. So you try to fine-tune it with real data, but… you never have enough real data, and the fine-tuning breaks everything the simulation taught it.

SimDist fixes this with one key decision: don't transfer the policy. Transfer the world model.

Keep the reward and value knowledge from simulation frozen. Only update the part that's actually wrong, how the robot predicts physics.

Now the robot doesn't have to relearn the entire task in the real world. It already knows what success looks like. It just needs to correct its understanding of how the real world moves.

The part that makes this work:

they also trained on failures and recoveries; not just perfect demonstrations. Without that, the planner finds the gaps and exploits them. With it, the robot can tell a good future from a bad one. That's all it needs.

Results on peg insertion, table leg assembly, locomotion on slippery and uneven surfaces. Tasks that require precision, force, and quick reaction.

Thanks for sharing, Tyler Westenbroek ([@ty_westenbroek].

Interactive visualization + paper: https://t.co/Ns4txW6Apk

——-

Weekly robotics and AI insights.
Subscribe free: https://t.co/9Nm01QUcw3

[Ryoichi Ando @ryichando]

We’ve released the official Blender add-on for ZOZO’s Contact Solver, our open-source physics solver developed at ZOZO, Inc. Blender users, please take a look!

https://t.co/t2kmcaCP2i

[Ronald van Loon @Ronald_vanLoon]

Reflex #Robotics Supercharges #Warehouses with Smarter, Faster #Automation
by @tweetciiiim

#Robots #Engineering #ArtificialIntelligence #Innovation #Technology

[Chris Paxton @chris_j_paxton]

Video action models are data efficient and allow robots to learn complex dexterous tasks. Learn more in our episode with @elvisnavah of @mimicrobotics ->

[James Wigglesworth @wiggleswor27016]

I don't think I've ever seen a gripper this versatile before - and I've seen a lot of crazy gripper designs.

[Lukas Ziegler @lukas_m_ziegler]

One tool to handle them all! 🧰

Traditional suction-based systems struggle with over 30% of SKUs. Mesh bags, loose apparel, and porous items often lead to failed picks or damaged goods.

Nexera Robotics (@NeuraGrasp) is pushing that ceiling to more than 95% by innovating the hardware, not just the software.

Their solution? NeuraGrasp®. Instead of rigid fingers or air-dependent suction, NeuraGrasp uses a compliant membrane that conforms to any object.

It creates a mechanical lock through friction and wrap-around contact, even on items that leak air or deform.

Pressure is distributed by the material itself, meaning no complex force-feedback loops are needed to prevent crushing. It’s single tool can lift heavy payloads, yet pick a single peach without bruising it.

A super important part is that no SKU-specific training and no tool changers required.

The industry has spent years optimizing robot “brains” while ignoring the hands. Nexera is perfecting the touch to make robots more capable.

Congrats Nexera Robotics team! 👏🏼

~~

♻️ Join the weekly robotics newsletter, and never miss any news → https://t.co/GoA3ZuwoPB

[Lukas Ziegler @lukas_m_ziegler]

BREAKING: Locus Robotics has acquired Nexera Robotics, and the technology they just bought could solve one of the hardest unsolved problems in warehouse robotics.

Grasping.

Not picking a box. Picking ANYTHING. Porous textiles. Loosely bagged items. Perforated polybags. Irregular packages. Delicate goods. Thin packaging. Items with inconsistent surfaces.

That's what Nexera's NeuraGrasp™ does, a single gripper that adapts dynamically to the physical characteristics of each item using AI-driven grasping intelligence, computer vision and a patented soft membrane structure.

The numbers behind it are serious:

Locus Array, already live in customer deployments now gets a manipulation capability that dramatically expands what it can autonomously handle.

The robots are coming.

@NeuraGrasp @LocusRobotics

[Chris Paxton @chris_j_paxton]

RLWRLD robot putting a phone in a box

from Eun Sook Moon on linkedin