Third Dimension

Third Dimension (Third Dimension AI, Inc) develops spatial reconstruction and generation technology that produces high-fidelity 3D environments from 2D images and video. Its product, SuperSim, is described as covering a "full fidelity spectrum," from reconstruction of real-world scenes captured from customer data through to open-ended generative simulation of imagined environments. The company positions the output as rendering-engine-ready content that can be plugged into an autonomy stack to train and evaluate embodied AI systems, with the stated goals of closing the gap between simulation and reality, improving data flywheels, and avoiding the cost of manually authoring scenes with large 3D art and engineering teams [0][2].

Technically, the company describes a large-scale 3D world reconstruction pipeline combining radiance-field methods (NeRF, Gaussian splatting) with diffusion-based image and video generation, converting a source image, video or blocked-out 3D layout into geometry that can be loaded into a real-time game engine. Demonstration material on the website includes scenes reconstructed from the WayveScenes101 dataset with faces blurred for privacy [0][7].

At launch the company framed itself broadly as a 3D generative AI company serving video games, film and virtual production, autonomous vehicles, drones, geospatial intelligence and military customers; later website messaging emphasizes autonomy teams and robotics simulation [5][7][0][2]. Publicly listed team members include CEO Tolga Kart, CTO Piotr Sokolski, Chief Researcher Christian Rupprecht, and staff across research, engineering, marketing and operations [2].