Robotics needs data. OVER delivers it. Robots aren’t just machines anymore, they’re becoming autonomous agents. The secret? A new generation of Physical AI Models. And the fuel? A global dataset built for scale. Let’s dive in 👇🧵

The Physical AI stack has 3 core layers: 👁️ Vision Models - The Eyes: he ability to perceive, localize, and reconstruct the 3D world with semantic understanding 🧠 World Models - The Imagination: the ability to simulate, predict outcomes, and plan actions in a safe virtual space 🗣️ VLA Models - The Hands & Ears: the ability to link language, vision, and action: turning human instructions into robotic execution These are the new primitives for building autonomous agents in the real world.

First up, Foundation Vision Models. 👀 Think of this as the robot's "visual cortex." It takes in camera data and builds a high-fidelity 3D map of its surroundings. It's how a bot understands physical space, depth, and objects. This is the perception layer.

Next, World Models. 🧠 This is the robot's imagination. It's a simulation engine in its head that predicts "what happens if I do this?" It understands physics and cause-and-effect, allowing it to plan complex tasks without breaking stuff. This is the strategy layer.

Finally, Vision-Language-Action (VLA) Models. 🗣️ This is the UI/UX layer. It connects natural language commands to what the robot sees and what it can do. You say "get me the cup," and the VLA generates the precise motor commands to execute it. This is how we'll talk to robots.

So here’s the catch: the entire stack depends on one thing: DATA. All these AI models need a massive, high-fidelity “textbook of the real world” to train on. The quality of the AI is only as good as the quality of its data, and OVER’s dataset is ultra-detailed. This is where a project like @OVRtheReality comes in. Our 3D mapping data is the fuel for this entire stack. → 153,000 3d maps of different locations of the world → ~77M high-res images → Rich RGB + Depth data This is a DePIN (Decentralized Physical Infrastructure Network) for world-scale AI training data.

Why is a DePIN for 3D data critical? Because no single company can map the entire world. A decentralized network of contributors can create a dataset diverse and massive enough to train robots that can operate anywhere, not just in a lab. This is how you scale real-world AI.

So the full picture looks like this: 1. A DePIN like OVER collects massive-scale 3D world data. 2. Vision Models train on this data to perceive reality. 3. World Models use this data to build predictive simulations. 4. VLAs use this context to turn human language into action.

This isn't just about cool robots. It's about building the foundation for autonomous economies, where agents can own assets and perform tasks in the real world. The intersection of DePIN, AI, and Robotics is where the next cycle's biggest narratives will be built. Read the full blogpost here: Read the OVER Wiki:

This is one of the clearest visions I’ve seen for bridging AI research and practical robotics. OVER’s approach to collecting 3D maps and depth data at scale is exactly what robots need to transition from labs to actual homes and workplaces.

X20 q4 on sight 🫡

OVER really feels like it’s sitting at the center of the robotics wave. If they can nail this global dataset, they won’t just power better robots.. they’ll define the standard for how Physical AI learns and scales.

Amazing the OVER technology is really wonderful and it is a real game changer

Robotics + AI = Future unlocked. OVER's dataset is the game changer! With 153,000 locations and 84TB of data, it's fueling the next gen of autonomous agents.

AI models eat data. Robots eat world data. OVER just turned the whole planet into a training set.

The use of AI is a very big advance along with robotics, this can really change the rules of the game, it is a big step, I will closely follow all these advances

OVER is solving a major bottleneck in robotics by providing large scale, high quality data. With Physical AI Models and a global dataset, robots can evolve into true autonomous agents that learn and operate in the real world.

This is wild. Robots are finally gonna understand the real world, not just a lab

Robots getting smarter by the day… OVER’s really feeding them the juice!

Really impressive work 👏 OVER isn’t just building smarter robots, it’s laying the foundation for real-world autonomous economies. Vision, world models and DePIN together feel like one of the biggest narratives of the next cycle

probably has to be one of the rarest i’ve seen with the OVER technology for me, it’s particularly the approach of collecting depth data and 3D maps

Physical AI layer is the real unlock here

Physical AI backed by real-world data feels like the unlock robotics has been waiting for.

Scaling AI + robotics takes two things: reliable data and a way for machines to interact economically. OVER covers the first, peaq covers the second.