What turns everyday human activity into something a machine can actually learn from? It’s not just the video itself. The value comes from how movement, context, timing, and interaction are captured and structured. That gap between raw activity and usable training data is where a lot of the real work happens. ZenO is focused on building that layer for Physical AI.

Go under the hood of Robotiq’s gripper simulation workflows in NVIDIA Isaac Sim. Join us and @Robotiq_Inc experts to learn how the team solved a complex articulation challenge using Newton, PhysX, and mimic joints. You’ll also learn about tactile sensors, deformable bodies, current Robotiq components, and what’s next. 📅 September 23, 2026 @ 11 AM PT 🔗 nvda.ws/4h6sINB

Missions #14 and #15 are now live. Mission #14 Refrigerator Organizing Take out, sort, move, and rearrange food and containers inside your refrigerator. Mission #15 Food Preparation Wash, cut, peel, measure, and prepare ingredients before cooking. Important: The Steps, Required, and Do Not sections have all been fully updated. Please read the mission guide carefully before recording and uploading. For both missions: - Head or face mounted phone only - Landscape mode required - Both hands must stay visible - First person POV only - 0.5x - 0.7x wide angle Upload via ZenO Core: app.zen-o.xyz Your data helps build the future of Physical AI.

ZenO Weekly Update Last week, we focused on expanding contribution options and strengthening the infrastructure behind ZenO. Here’s what changed: - New Voice Mission is live: Tell Us About Your Day. Contributors can now submit short English voice recordings directly through ZenO. - We strengthened monitoring in Sim Teleoperation to detect abnormal activity and restrict suspicious bot-driven accounts. - We improved parts of our backend processing and validation pipeline to better handle incomplete submissions, failed processing jobs, and anomalous data before rewards are calculated. - Additional data integrity checks are being introduced to make validation more consistent across different contribution types. - We continued improving system reliability as submission volume grows across ZenO. Cleaner data, stronger validation, and more reliable infrastructure behind every contribution. More updates soon.

Building ZenO isn’t just about collecting more data. We’re continuously improving missions, validation, contributor tools, and the pipeline behind every submission. There’s more being built behind the scenes. More updates soon.

New Voice Mission is live: Tell Us About Your Day Talk freely about what happened today for 1–2 minutes. No script just speak naturally, like you’re talking to a friend. For the best recording quality: • Record in a quiet room • Keep your device at least 10 cm away • Use the device microphone, not a headset or earphones • No music or other audio in the background • If someone else is recorded, make sure they’ve agreed to participate Start recording on ZenO: app.zen-o.xyz/data/audio

Real-world data for Physical AI. ZenO contributors have captured 400K+ minutes of egocentric real-world activity across everyday environments and manipulation tasks. But it’s more than video. Our capture pipeline is designed to collect multimodal signals that help robots understand how humans interact with the physical world: • Egocentric RGB video • Depth data • Head trajectory and camera motion • Hand motion and interaction context • Real-world object manipulation across diverse environments Data is captured through smartphones and wearable devices, then structured for robotics and embodied AI training workflows. The goal is to move beyond passive video datasets toward richer human interaction data that can support perception, action understanding, and robot learning.

🧹 Mission #4 is back! Turn everyday organizing into valuable real-world data for Physical AI. Tidy up your room, organize storage, or put items back in place and contribute to the future of embodied AI. Join Mission #4 today.

The real world is training Physical AI. Start contributing with ZenO: app.zen-o.xyz/

Something is cooking behind the scenes 👀 We’re preparing something new for the ZenO ecosystem. More soon.

3 months of ZenO Lab. Since June 1: • 216K+ robot manipulation attempts • 202K+ successful attempts • 199K+ trainable trajectories • 1,896 hours of successful manipulation data • 20 robot tasks • 70 policy training and evaluation runs For context, the original LIBERO benchmark contains roughly 20,000 episodes. ZenO Lab runs directly in the browser with no installation or physical robot required. Users generate robot trajectories, success is validated automatically, and new data is fed into a daily training and evaluation loop. Recent policies have reached up to 100% success, with several tasks peaking in the 80–95% range. Our latest rendered dataset includes: 1,293 episodes 649,590 frames ~15GB of training data Exportable in LeRobot v3 format. ZenO Lab is becoming infrastructure for turning human-generated robot experience into continuously improving robot policies. And we are only 3 months in.

ZenO Weekly Update Last week, we continued expanding how real-world and simulated data are collected and shared across the ZenO ecosystem. Here’s what changed: - ZenO datasets are now live on Hugging Face, including real-world egocentric data, GoPro-based first-person data, and Sim Teleoperation datasets. - Mission #13: Tie Your Shoelaces is now live, adding another everyday first-person activity to our real-world data collection missions. - A new REAL WORLD mission type has been introduced to ZenO Lab, bringing rendered everyday objects into the sim teleoperation environment. - REAL WORLD missions also receive a higher XP multiplier, rewarding contributors for completing more realistic simulation tasks. We’re continuing to connect real-world human activity with simulation to build useful training data for Physical AI.

🌎 New ZenO Lab Mission: REAL WORLD Real-world objects are entering ZenO Lab. REAL WORLD adds rendered everyday objects to our sim teleoperation environment, bringing each task closer to how robots interact with the physical world. And there’s more: REAL WORLD missions receive a higher XP multiplier. Teleoperate. Complete missions. Earn more XP. Generate data for Physical AI. Enter ZenO Lab and try REAL WORLD now. lab.zen-o.xyz/missions

Mission #13 is now live. Record yourself naturally tying your shoelaces. - 5 to 20 minutes - Head or face mounted phone only - Landscape mode required - Both hands must stay visible in frame - First person POV only - 0.5x - 0.7x wide angle - Upload via ZenO Core: app.zen-o.xyz - Discord: discord.com/invite/5u5626MYP… Your data helps build the future of Physical AI.

Physical AI starts with data. More physical data leads to stronger generalization.

ZenO Weekly Update Last week, we focused on improving how real-world data is captured and rewarded across the ZenO ecosystem. Here’s what changed: - ZenO App V2 is live with Profile Cards and Voice Missions, available on both iPhone and web. - iPhone app contributors now receive 30% more XP for eligible data submissions. - We shared a closer look at ZenO’s ARKit-based data pipeline, capturing not only RGB video but also 6DoF trajectories, spatial information, hand movement, and LiDAR-based depth. - A new iPhone Contributor role has been added to recognize active contributors using the ZenO app. We’re continuing to improve the way real-world human activity is turned into useful training data for Physical AI. More updates coming soon.

Newton 1.5 is now available, making robot training at scale faster and more reliable. This release delivers: 🔁 More parallel simulation, lower memory, selective resets 🤝 Consistent contact physics 🤖 Experimental batched GPU control 📦 Cleaner USD/MJCF imports Get started: github.com/newton-physics/ne…

ZenO App V2 for iPhone is officially live. This update introduces Profile Cards and a new Voice Mission feature. Voice Missions are also available directly through app.zen-o.xyz, so you can participate from both the app and the web. With V2, ZenO is evolving beyond simple video capture into an app that can generate robot-learning-ready data at the moment of recording. ZenO uses ARKit on iPhone and will use ARCore on the upcoming Android app to capture camera motion and spatial information, including 6DoF trajectories, alongside video. On LiDAR-supported devices, ZenO can also capture absolute distance and depth information without requiring separate post-processing. Instead of recording ordinary video first and converting it later, ZenO is designed to create data that is immediately more useful for Physical AI and robotics training. ZenO for Android is coming soon. 👉apps.apple.com/app/zeno-data…