Learn About Odyssey's AI Products
Create a free account to access in-depth lessons on each tool and model.
Start Learning FreeπAbout Odyssey
Updated June 18, 2026Odyssey is an AI startup, founded in 2023 and headquartered in San Francisco, building learned world models β generative simulators that learn how interactive environments evolve from data rather than from hand-coded game engines. It is led by CEO Oliver Cameron, previously a co-founder of self-driving startup Voyage (acquired by Cruise) and an engineering leader at Cruise itself; CTO Jeff Hawke was an engineer at UK self-driving company Wayve. Both bring deep experience in real-time generative systems for autonomy.
The technical bet is that a world model can serve as a learned game engine β a single neural system that handles both how the world evolves in response to actions and how it appears visually, without the traditional separation between physics simulation and graphics rendering. Odyssey's flagship system Agora-1, introduced in May 2026, demonstrates this by letting up to four humans or AI agents share the same real-time generated simulation. The architecture decouples simulation from rendering: one learned function evolves the world state based on player actions, while a diffusion-transformer-based model renders each player's view from that shared state.
The May 2026 launch demo runs as a GoldenEye-style four-player deathmatch β a deliberately consumer-recognizable format meant to make the multi-agent world model visible to non-specialists. Beyond gaming, Cameron has flagged robotics, defense, education, foundation-model research, and multi-agent reinforcement learning as longer-term applications: any domain where multiple agents need to share a real-time simulated environment that updates in response to their joint actions. Cameron framed the release as "an entirely new class of interactive systems."
Agora-1 is available for testing at agora.odyssey.ml. In 2026 Odyssey raised a $310 million Series B at a $1.45 billion valuation, led by Natural Capital with participation from Amazon, AMD Ventures, and Google's GV β bringing total funding to about $337 million. As part of the Amazon backing, AWS becomes Odyssey's preferred cloud and the company will optimize its models for Amazon's Trainium chips. The company's positioning bet is that learned world models will sit between today's video-generation systems (one-direction generation, no interaction) and tomorrow's full embodied-AI simulators (rich interaction, physical fidelity) β a category that DeepMind's Genie, NVIDIA's GR00T research, and OpenAI's Sora work have all gestured toward without converging on a shared interface for multi-agent use.
π οΈProducts & Tools (1)
Learned world model from Odyssey that supports real-time multi-agent shared simulation β up to four humans or AI agents in the same generated environment. Decouples simulation from rendering via a diffusion-transformer architecture. Launched May 19, 2026; targets gaming, robotics, defense, and multi-agent reinforcement learning research.
