🚀Introducing Ropedia Xperience 2
What does it take to scale Physical AI? We believe it takes both *environments* and *behaviors*
🏠1,589 metric-scale 3D scene scans
🎥4,230 paired egocentric recording
🤖74 tasks across 131 hours of human demonstrations
https://t.co/TohrmWPPf5
@ropedia_ai From this capture we build simulation-ready scenes, robot trajectories and long-horizon task data, with no robot on site.
The release: 1,500+ scenes, 4,000+ recordings, 74 tasks and three benchmarks.
Introducing Xperience 2 from @ropedia_ai : a gold standard for scaling Physical AI environments and behaviors ! 🎉
📄 Tech report: https://t.co/ZSN0xPlZ6s
🌐 Project page: https://t.co/AUYd0pvhAF
🤗 Dataset: https://t.co/4Y7N6Kgt07
Operators record first-person video together with hand and full-body motion while doing everyday tasks in real scenes. Every scene is also 3D scanned, and each recording is aligned to its scan, so the video, the motion and the 3D space share one coordinate frame.
Introducing S-Agent 2.0, outperforming GPT-6 Astra on MMSI-Bench and ReVSI-Bench!
A general framework for evidence-grounded spatial intelligence that turns coding agents into spatial reasoning agents.
📄 https://t.co/t6rHBnERW3
🌐 https://t.co/FM4uqsL2RD
We’ll also release HoliWorld-Data, featuring 100 scenes and 400+ human egocentric videos, enriched with holistic multimodal annotations automatically generated by the HoliWorld pipeline.
Thrilled to introduce HoliWorld, developed at @ropedia_ai! 🌍
The first scalable auto-labeling pipeline that transforms egocentric videos into holistic multimodal data from 3D bounding boxes to 4D object tracking and beyond.
🤗 https://t.co/RXRcyVo1j1
🌐 https://t.co/gay22f2bbL
Our ancestors picked up a rock. They saw a hammer, a blade, and a way to survive. Humans didn’t conquer the world with bare hands -- we did it with tools. 🛠️
Behind every tool we reach for is a tiny act of intelligence: "What do I need? What should I use? What do I do next?"
It's not just knowing what something IS. It’s imagining what it could DO. Can current AI models reason the same way?
Meet #EgoTools , collab with @ropedia_ai.
🔗 https://t.co/STbOC1dWuL
Are video models Aristotle—or Newton? 🍎
Do they actually reason with physical laws, or only generate motion that looks plausible?
Apple-π is the first law-grounded benchmark: 400 videos, 10 mechanics tasks, Perception→Formulation→Deduction.
🌐 https://t.co/sE04Arxqji
Or… what if we gave you $100 in Codex credits if you tell us what you love about GPT-5.6 Sol or why you switched?
Tweet it, claim your gift, enjoy more usage. First 10k get the free tokens!
https://t.co/8mU93eA13i
We finally release S-Agent 🎉
The first agentic model for Spatial Intelligence.
S-Agent turns perception into action: grounding, reconstructing, and reasoning with tools to solve complex spatial tasks step by step.
Page: https://t.co/x62xRkqxiX
Paper: https://t.co/vnEWb2Qsft
Real-world intelligence starts with real-world data.
We’re bringing Xperience-10M to @CVPR — helping build the data foundation for the next generation of Physical AI.
Looking forward to connecting with researchers, builders, and industry leaders shaping the future of AI.
#CVPR2026 #PhysicalAI #EmbodiedAI #ComputerVision #Ropedia
What makes a truly reliable spatial benchmark?
Today, Ropedia is excited to introduce SpatialBench — a comprehensive benchmark framework for Spatial Intelligence.
SpatialBench evaluates 42 models across:
• 5 domains
• 19 datasets
• 500+ real-world scenes
Spanning egocentric, robotics, indoor, and outdoor environments.
Alongside SpatialBench, we’re also introducing DepthAnything-Next — a vision model designed for stronger robotics and ego-centric understanding.
Our paper is now ranked #2 on @huggingface Daily Papers.
As Spatial AI moves toward embodied and real-world systems, we believe evaluation infrastructure will become foundational to the next generation of spatial intelligence.
More In:
https://t.co/hXg7vFiTg5
https://t.co/wmjQsWsFiO
#SpatialAI #EmbodiedAI #SpatialIntelligence #ComputerVision #Robotics #WorldModels #AIResearch #OpenSourceAI #3DReconstruction #PhysicalAI
First work done in @ropedia_ai 🎉
🔥SpatialBench: Is Your Spatial Foundation Model an All-Round Player?🔥
We evaluate over 19 source datasets, 540+ scenes, 40+ model variants, and 6 reconstruction paradigms.
More In:
https://t.co/YAUCixMw1H https://t.co/vccqXSHhjJ
A privilege to brief Mr. Heng Swee Keat 🇸🇬 on @ropedia_ai latest developments today, alongside our CEO @Frozen_Burning and CTO @hongfz16 ! 🚀🚀
Incredibly proud to be part of the founding research scientist team behind this journey. Onward and upward! ✨📈
Ropedia team @Frozen_Burning@hongfz16 spoke with Former Deputy Prime Minister Heng Swee Keat of Singapore🇸🇬 about physical intelligence and real-world data at the launch of NTU × UOB Innovation Hub. Honoured to be invited by @NTUsg.
More to come.
#EmbodiedAI#PhysicalAI#AI #Deeptech #Singapore #Startup