Releases·Model Releasefeatured
World Labs releases Atlas, a spatial world model generating unseen viewpoints and navigable space-time simulations
World Labs announced Atlas, a foundation model for spatial intelligence that ingests text, images, image sequences, camera poses, and depth to build persistent spatial understanding of scenes. Built on an autoregressive diffusion transformer, it generates unseen viewpoints and infers missing geometry rather than only reconstructing observed content, and converts scenes into Gaussian splats for fast rendering. From footage captured on just 3-5 consumer phones it can build complete navigable space-time simulations, positioning it as a step toward physical AI.
386 engagement3 sourcesTue, Sep 1, 2026, 07:25 PM
AI Overview
- Atlas is an autoregressive diffusion transformer that grounds text, images, video-like sequences, camera poses, and depth in a shared 3D "spatial context" instead of treating video as 2D pixel grids
- It can generate unseen viewpoints and infer missing geometry beyond what was directly observed, producing practical 3D representations like Gaussian splats for fast rendering
- From footage of 3 to 5 standard cell phones, Atlas builds a complete, navigable space-time simulation — a democratized "Bullet Time" capability
- Real-to-Sim features that turn captured footage into physical simulations are highlighted as key to accelerating physical AI
- Coverage frames the release as moving AI past basic generation into true physical simulation
Related
Social·Community
World Labs' Atlas claims 50-100x faster 3D scanning from photos, but Ben Mildenhall calls reliability a fundamental challenge
A Reddit analysis of an a16z clip featuring Fei-Fei Li and Ben Mildenhall discusses World Labs' Atlas model, which infers 3D geometry from plain photos and claims 50-100x cost savings over manual scanning. The author, citing a construction background, compares Atlas with LiDAR and argues LiDAR keeps an edge where measured accuracy matters. Mildenhall reportedly acknowledged that making the product reliable remains a fundamental challenge.
A Reddit analysis of an a16z clip featuring Fei-Fei Li and Ben Mildenhall discusses World Labs' Atlas model, which infers 3D geometry from plain photos and claims 50-100x cost savings over manual scanning. The author, citing a construction background, compares Atlas with LiDAR and argues LiDAR keeps an edge where measured accuracy matters. Mildenhall reportedly acknowledged that making the product reliable remains a fundamental challenge.
0 engagement1 sourceredditThu, Sep 17, 2026, 11:16 AM
News·Company Newsfeatured
AMD will acquire Fei-Fei Li’s World Labs for $8.2 billion
AMD announced a definitive agreement to acquire World Labs, Fei-Fei Li's spatial-intelligence startup, for approximately $8.2 billion in an all-stock deal, its second-largest acquisition after Xilinx. Li joins AMD as EVP and chief scientist reporting to CEO Lisa Su, with co-founders Justin Johnson and Ben Mildenhall leading the team inside AMD. AMD says World Labs' frontier workloads will shape its chip roadmap, strengthening its push against Nvidia in world models for robotics and simulation.
AMD announced a definitive agreement to acquire World Labs, Fei-Fei Li's spatial-intelligence startup, for approximately $8.2 billion in an all-stock deal, its second-largest acquisition after Xilinx. Li joins AMD as EVP and chief scientist reporting to CEO Lisa Su, with co-founders Justin Johnson and Ben Mildenhall leading the team inside AMD. AMD says World Labs' frontier workloads will shape its chip roadmap, strengthening its push against Nvidia in world models for robotics and simulation.
240173 engagement7 sourcesrss, redditMon, Sep 28, 2026, 08:39 PM
Releases·Model Release
Runway's GWM Worlds 2 Adds WorldPrompt for Real-Time Interactive World Generation
Runway introduced GWM Worlds 2, a research preview that turns high-fidelity video and audio generation into real-time interactive simulation, and detailed its new WorldPrompt input format in interviews with CTO Kamil Sindi, Principal Research Scientist Robin Kahlow, and co-CEO Anastasis Germanidis. WorldPrompt lets users fix parts of a simulated environment and prompt timestamped, real-time events, differentiating Runway from Genie 3, Odyssey-2 Pro, and World Labs' RTFM. The work matters to practitioners because real-time autoregressive world generation remains an open engineering problem with limits in error accumulation, memory, and counterfactual realism.
Runway introduced GWM Worlds 2, a research preview that turns high-fidelity video and audio generation into real-time interactive simulation, and detailed its new WorldPrompt input format in interviews with CTO Kamil Sindi, Principal Research Scientist Robin Kahlow, and co-CEO Anastasis Germanidis. WorldPrompt lets users fix parts of a simulated environment and prompt timestamped, real-time events, differentiating Runway from Genie 3, Odyssey-2 Pro, and World Labs' RTFM. The work matters to practitioners because real-time autoregressive world generation remains an open engineering problem with limits in error accumulation, memory, and counterfactual realism.
0 engagement1 sourcerssFri, Sep 25, 2026, 01:30 AM
Developments·Paper
Agentic Real2Sim framework automates real-to-sim conversion for robotics
Researchers introduced Agentic Real2Sim, a framework that uses vision-language agents to automate the labor-intensive process of converting real-world scenes into physics simulations. The system recovers scene geometries, object states, and physical parameters, then assembles actors, objects, cameras, poses, and trajectories into a runnable simulation, eliminating manual tuning and brittle workflow glue.
Researchers introduced Agentic Real2Sim, a framework that uses vision-language agents to automate the labor-intensive process of converting real-world scenes into physics simulations. The system recovers scene geometries, object states, and physical parameters, then assembles actors, objects, cameras, poses, and trajectories into a runnable simulation, eliminating manual tuning and brittle workflow glue.
0 engagement1 sourcearxivTue, Jul 21, 2026, 03:23 PM
Developments·Paper
Researchers use LLMs as spatial software generators for programmable 3D objects
A new paper, shared by a co-author on Reddit, explores using LLMs to generate 3D objects as spatial software rather than monolithic meshes. The approach yields objects composed of logical parts that are inherently programmable and animatable. Demonstrations are available at nova3d.xyz.
A new paper, shared by a co-author on Reddit, explores using LLMs to generate 3D objects as spatial software rather than monolithic meshes. The approach yields objects composed of logical parts that are inherently programmable and animatable. Demonstrations are available at nova3d.xyz.
0 engagement1 sourceredditMon, Aug 24, 2026, 07:10 PM