The wait is over. HOMIE Gen2 is here. Lightweight. Ergonomic. 360° vision. Spatial audio. Precise sync. Training-ready multimodal data, precisely captured and synchronized. Unlock the Experience Scaling Law for Physical AI — with HOMIE Gen2.

4/5 The Results Can a world model help evaluate robot behaviours before real-world deployment? In reported experiments: • Ropedia W0-predicted policy scores achieved a 0.996 correlation with real-robot results. • Model-based planning achieved 55% task progress, compared with 26% for random selection and 45% for the best fixed policy. These results highlight the potential of world models for robot policy evaluation and planning.

3/5 Scaling Experience Ropedia W0 is powered by: • 10K hours of human experience • 3K hours of robot experience By learning from both human and robot data, Ropedia W0 explores how physical experience across different embodiments can support more generalizable world models.

2/5 The Core Idea Introducing Ropedia W0: Scaling Loco-Manipulation World Models across Embodiments with Human Experience. W0 predicts future first-person videos from human or robot motion instructions, exploring a shared approach to learning across humans, robotic grippers, and humanoid robots.

📣 Ropedia Super Week | Oct 19–23 Ropedia Super Week starts tomorrow. Let's go! The full line-up: Oct 19: Ropedia W0 Our world model Ropedia W0 learns from human video to model walking and dexterous manipulation across humans and robots. Oct 20: HoliWorld + EgoXperience HoliWorld connects scenes, object motion and interaction histories to create training questions and answers backed by visual and geometric evidence. EgoXperience spans data capture, annotation, learning and evaluation for joint 3D reconstruction from first-person video. Oct 21: S-Agent 2.0 A spatial reasoning agent that connects observations, measurements and persistent scene memory to answer 3D questions from images and video. Oct 22: DONIC A generalist controller that learns from human demonstrations to coordinate two dexterous hands. Oct 23: Ropedia A0 A robot recursive self-improvement (RSI) framework in which coding agents write, run and revise reusable code policies offline in simulation. First up tomorrow: our world model Ropedia W0. ✨

Ropedia Super Week | Oct 19–23 6 breakthroughs. 5 days. A showcase of how first-person human data can advance embodied AI, from spatial reasoning to action and control. Here's the lineup. ↓ Mon: Ropedia W0, our latest robot world model. Tue: HoliWorld + EgoXperience, our spatial data pipeline & full-stack data-to-model framework. Wed: S-Agent 2.0, our evidence-driven spatial reasoning agent. Thu: DONIC, our generalist controller for two dexterous hands. Fri: Ropedia A0, our code-driven robot self-improvement framework. Tomorrow, we'll take a closer look at our projects. Follow Ropedia for a first look.

That’s a wrap at IROS—but the conversations continue! A huge thank you to everyone who stopped by our booth. Didn’t get a chance to explore our IROS evaluation package? Get in touch with our U.S.-based team to learn more and explore collaboration opportunities.

Day 1 at IROS is wrapped—and Day 2 is already in full swing! A big thank-you to everyone who came by yesterday and spent time with Ropedia and Homie Gen 2. We’re back at Booth 742 today, come say hello and give Homie Gen 2 a try.

Our booth at IROS has been buzzing with visitors! It’s been great welcoming everyone to Booth 742 to explore Ropedia and try Homie Gen 2 firsthand. Thanks to everyone who stopped by—we look forward to meeting more of you at IROS! #IROS #Ropedia #HomieGen2 #Robotics

Physical AI starts with real-world data. Meet Ropedia at #IROS2026: → HOMIE Gen2 live demo → IROS-exclusive Evaluation Dataset → Ropedia merch 📍 Booth 742 📅 Sep 28–30 Come see what we’re building.

Today, we’re pleased to announce the Ropedia Academic Partner Program, supporting researchers advancing Physical AI, robotics, and embodied intelligence. Academic partners will gain access to HOMIE Gen2 to capture real-world multimodal data and explore new possibilities in robot learning and embodied AI. We believe the future of Physical AI will be built together across academia and industry. Join us today and let's build together: ropedia.com/collaboration

From research to building for the real world. HOMIE Gen 2 is our next step toward making real-world human experience accessible to the next generation of robots and Physical AI systems. Proud to have our journey featured by NTU, and excited for what’s ahead. 🚀

Hello from Malmö! 🇸🇪 What a day at PhysAI 2026 @eccvconf 2026! A packed room. 10 invited talks. Researchers from Oxford VGG, Stanford, MIT, ETH Zurich, NVIDIA, Google DeepMind, UC Berkeley & more. One shared focus: advancing Physical AI. Proud to support the community building the foundations of physical intelligence. 🌍

AI scaled with data. Physical AI will scale with experience. Tomorrow. Sharper capture. Deeper sensing. Tighter sync. HOMIE Gen2.

Today, Ropedia announces US$30M in total funding to scale the data infrastructure empowering Physical AI. More real-world data. More experience. Better Physical AI. We’re just getting started.

Congratulations to @Alibaba_Qwen @Ali_TongyiLab on Qwen-VLA, a unified vision-language-action model spanning manipulation, navigation, and trajectory prediction across robot embodiments. We're proud that our Xperience-10M is one of the egocentric human datasets behind it, contributing real first-person experience with depth, hand and body motion, and language. Physical AI is only as good as the human experience it learns from — and that's exactly what we're building Xperience for. Check our Xperience-10M: huggingface.co/datasets/rope… #PhysicalAI #EmbodiedAI #Robotics #VLA

Meet S-Agent, the latest from Ropedia — spatial reasoning as an agentic loop, not a single-shot answer. S-Agent acts to gather evidence: • a VLM planner decides what evidence is missing • tools handle 2D grounding, 3D lifting, measurement, orientation • scene + agent memory carry evidence across views and tool calls Zero-shot, it reaches 46.4% on MMSI-Bench, 60.0% on ViewSpatial-Bench, and 58.8% on ReVSI. Distilled into S-Agent-8B, it lifts Qwen3-VL-8B from 31.1% to 41.6% on MMSI-Bench and 46.8% on ViewSpatial-Bench. - Page: ropedia.github.io/S-Agent - Paper: arxiv.org/abs/2606.20515 - Code: github.com/Ropedia/S-Agent #SpatialAl #SpatialIntelligence #PhysicalAI #EmbodiedAI