World Labs, founded by Fei-Fei Li, unveiled Atlas, an AI model that generates complete 3D scenes and minute-long 1440p videos from a single photograph. Atlas uses a "world model" approach grounded in spatial reasoning—understanding 3D structure rather than just predicting the next frame—and operates as an omni-model processing text, images, video, and 3D data in one architecture. With $1.23B raised and $5B valuation, Atlas addresses robotics, gaming, design, and VFX needs by outputting usable 3D geometry, requiring only 2–3 input photos for accuracy while managing 100+ images for high precision. This positions it as a significant step forward in novel view synthesis and spatial AI.