Dear users, we would like to inform you that the technical difficulties experienced earlier have been resolved.

All bot services have resumed normal operation, and all functions are accessible. We appreciate your understanding.

Bots:
https://t.me/neuralpony/3

Robots learning in simulation, and then mastering the real world.

[📍 Bookmark GitHub Repositories]

Today’s video shows a full Sim2Real pipeline with the OMY robot.

✅ Compact 6-DoF manipulator built for AI and robotics research
✅ Trained in Isaac Sim, validated in Gazebo, deployed in reality
✅ Ideal for imitation learning, RL, and Sim2Real experiments

Watch the full video and see the complete workflow in action.

AI Manipulator and Open Manipulator
https://github.com/ROBOTIS-GIT/open_manipulator

robotis_lab
https://github.com/ROBOTIS-GIT/robotis_lab

Resources:
https://forum.robotis.com

Researchers at the National University of Singapore have created a new underwater robot inspired by one of the ocean’s most intelligent creatures - the octopus. Designed for both practical underwater tasks and biomechanical research, the robot mimics the octopus’s unique swimming and movement abilities.

“Our research journey began long ago with studying diverse principles for reaching and grasping, locomotion, and swimming,” said Prof. Cecilia Laschi, who led the study.

Unlike many marine animals, octopuses are exceptionally agile - they can swim swiftly and interact with objects around them. The goal of the study was to develop a robotic system that closely replicates these advanced underwater capabilities.

A Canadian ag-tech startup, Mycionics, has built a robotic system that aims to bring more innovation into mushroom harvesting. 🇨🇦

Their setup uses dual-arm gantry robots, machine vision, and soft-grip tools to pick mushrooms.

The system doesn't replace humans entirely, but works collaboratively, handling repetitive tasks like picking, packing, and sorting, while also scanning each mushroom for real-time analytics.

It integrates with existing farm setups and scales easily. Already rolling out in North America, with European expansion underway.

By improving consistency and boosting yields up to 8%, Mycionics tackles labor shortages and food safety-showing how robotics can elevate even the most delicate, niche farming tasks.

Mushrooms, be ready, robots are coming for ya! 🍄

Thoughts? 🤔

"If you talk to ChatGPT about your most sensitive stuff and there's a lawsuit, we could be required to produce that..."

- Sam Altman, CEO of OpenAI

Chat: @neuralforum

👦🏼: ‘Mom, can we have Optimus Prime?'

👩🏻: ‘No, there is Optimus Prime at home…’

Optimus Prime at home 👇🏼

Creating something cool with robotics requires building it first.

So, the DIY aspect of robotics is essential.

Servos, cardboard boxes, and a simple controller, on which the control system is based, can be used to build a transformer robot....at home! 🏡

Beautiful moments of humanoid robots and humans. Humanoid robot (Astribot S1) is the music conductor, directing the orchestra and the audience to complete a great performance. Atmosphere, emotion, response, movement, collaboration, experience...

2024: everyone releasing their own Chat
2025: everyone releasing their own Code

Figure 02 folds laundry using the Helix AI model developed by Figure.

Figure claims this is the first instance of a humanoid robot with multi-fingered hands folding laundry fully autonomously using an end-to-end neural network.

Helix processes vision and language inputs to deliver precise motor control, handling deformable objects like towels without explicit object representations.

This also showcases adding new tasks only with training data, without requiring architectural changes to the model.

Disarm, grow, change, and start over

NeuralOS: a generative OS entirely powered by neural networks

NeuralOS, probably the world's most expensive operating system, running at a blazing 1.8fps on an NVIDIA H100 GPU. 😅

What exactly is NeuralOS?

It's an experimental generative OS that predicts every screen frame entirely from your mouse and keyboard inputs. No internet, no traditional software stack, purely hallucinated pixels.

How does it work?

- An RNN tracks the computer state (kind of like a traditional OS kernel, but all neural and continuous).
- A diffusion model generates the actual screen images (imagine a desktop environment, but fully neural-rendered).

The GIF shows a funny demo: NeuralOS running NeuralOS inside itself. Every single pixel you're seeing is model-generated, no network involved at all!

Long-term, our goal is to remove boundaries between software entirely and make OS fully customizable beyond fixed menus and options. Imagine asking your OS something like:

- "Merge all my messaging apps into one interface."
- "Make Signal look like Messenger."
- "Turn the movie I'm watching into a playable video game."

I'm curious about your thoughts:

Could future OS interfaces just become human-like avatars (think Grok's Ani)? Are menus and app-specific UIs going away?

What about fully generative games: could diffusion-based games eventually replace traditional ones?

Try the live demo here: neural-os.com (you might need patience…)

Logo reveal in ink-in-water style with VEO3 🖋💧

Runway Aleph can accurately remove subjects from your scene. Even with complex motion, lighting conditions and passing foreground occlusions. All you need to do is tell Aleph what you want.

Update in @neuralforum Chat

We are pleased to announce that a new image generation model, GPT-4o, is now available.
Other models such as DALL·E 3 and Google Gemini Images remain accessible as well.

How to use:

/o prompt → Generate images with GPT-4o

/i prompt → Generate images with DALL·E 3

/g prompt → Generate images with Google Gemini

👉 The bot is available in this topic:
https://t.me/neuralforum/7333097

✅ All requests are free and unlimited.