AI learns from data. Physical AI learns from the real world. At ZenO, we’re building the foundation of Physical AI with real-world human data. 🌍

“PrismaX is helping set a new standard for community-driven data quality validation. We’re excited to be part of this milestone as a Day 1 Launch Partner and look forward to seeing the ecosystem grow.” - @ZenO4AI @ZenO4AI is a physical AI data network, collecting first-person multimodal data for robotics and embodied AI. They share our belief that robots learn best from a human perspective, and were among our most active testers.

Applications are coming in from around the world. 📱 Applications are still open. If you have an iPhone 12 Pro or newer, apply now to join the ZenO iPhone Contributor Program. 👇 forms.gle/NYXe7UPVefEgXH9M9

Home robots are coming. 🤖 And behind them lies an enormous need for real-world data. 👀

Hiring specialized operators and setting up capture rigs that cost tens of thousands of dollars per unit runs into a clear limit. It is slow, it is expensive, and it never spreads widely. The iPhone already sits in billions of pockets. We believe it is the most realistic key to breaking through that ceiling.

An iPhone, with its LiDAR and precise sensors, can actually capture this 3D information that web video has never been able to provide. And this is not simply a question of better data. It is a question of scale. To make robots genuinely useful in the real world, you need a vast amount of diverse data. And to push diversity and scale, the tool doing the capturing must never be complicated.

What a robot truly needs is a completely different kind of data. The distance to each object in the frame, how that object moves, the friction of its surface, its inertia, and the 3D world coordinates traced by human hands and arms through space. Data that lets a machine understand the world as a three dimensional physical space rather than a flat image.

Building the hardware itself is, surprisingly, not the hardest part. The hard part comes next. Getting a robot to pick up a screwdriver off the table, line up the angle, and drive in a screw. Something a three year old can do is still a massive wall for a robot. And that wall comes down to data. The video data we usually talk about, the endless clips flooding YouTube, is nothing more than a flat 2D picture to a robot. Just an array of pixels, with almost none of the information a robot could actually act on.

Why did we build an iPhone app? To answer that, you first have to see the robotics industry for what it really is. People think of robotics as a hardware industry. Motors, actuators, joints, metal frames. But the closer you get to the actual work, the more you realize it is almost the opposite. Robotics is far more software driven than most people assume. So much so that you could say a robot is only as good as its data.

Do you own an iPhone Pro? 👀 Applications for the ZenO iPhone Contributor Program are now open. Selected participants will contribute to real-world AI data collection and receive: - Exclusive Discord iPhone Contributor Role - Separate Token Rewards Requirements - iPhone 12 Pro or newer Apply now 👇 forms.gle/NYXe7UPVefEgXH9M9

ZenO App 📱 Loading... 98% ⏳ Almost ready. See you soon.

We’re excited to join @PrismaXai as a Day 1 Launch Partner. We’re proud to be part of this important milestone and look forward to seeing PrismaX continue to grow.

The First 100 begins now. Verify Quality is live on PrismaX. For the first time, anyone can score the robot training data that models learn from, earn points, and compete to become one of The First 100. Better data. Better models. The standard starts with you.

ZenO is now ranked Top 8 on the @base App Leaderboard. Building the data network that connects Physical AI and onchain. Growing on Base. 🚀

This is exactly why ZenO collects data directly. We build first person data where the 6DoF trajectory of the hand and clear action labels are aligned from the very start. That is a different starting point from data scraped off the web and patched with guesswork after the fact. There is a lot of video on YouTube. There is still very little data a robot can actually learn from. app.zen-o.xyz

The second problem is the embodiment gap. A human hand and a robot gripper are built differently. You can hand a robot a video of a person pinching a cup with their fingers, and it still cannot map that motion onto its own body. Human motion does not translate directly to a robot's kinematics or control interface. There is a gap between the action in the video and the action a robot actually executes that is hard to close. The third problem is unrefined context. Web video constantly changes viewpoint, gets edited, and is full of off task segments. It is not cut into training ready episodes, and the alignment between narration and the actual movement is loose, which pollutes the learning signal itself.

For a robot to learn a movement, it needs to know where the hand went at every moment, and along what trajectory. But a YouTube video does not contain that trajectory, the robot executable action label. Video shows you the outcome. The information of "what action was performed" is written nowhere in it. The research community is explicit about this: human videos are not action data for robots, and they lack aligned, robot executable action labels and proprioceptive context. So researchers run hand pose estimators, bolt on depth models, and mix in separate robot data to try to recover that trajectory after the fact. They are filling in missing information with guesswork, and the noise from that process piles up.

"YouTube already has millions of videos. Why not just use those?" It's a fair question. We get it a lot. YouTube has millions of hours of people cooking, cleaning, and assembling things. It's free, diverse, and massive in scale. So we didn't dismiss it. We took it seriously and ran it all the way through. There is real, active research trying to scrape web video into training data. Our conclusion did not change. And the biggest reason was not the one we expected going in. The most fundamental problem is that web video is not "action data."

Discord XP is still live. 🚀 Chat, contribute, and stay active to earn XP and level up. Higher levels = stronger TGE rewards. Join the community: discord.com/invite/5u5626MYP…