The Generative Reality: Why We’re Investing in Odyssey (Copy) (Copy)
Since the dawn of digital entertainment, we have lived in a world of "baked" pixels.
Whether it is a Hollywood blockbuster, a AAA video game, or a social feed, the digital experiences we consume are static artifacts. They are pre-rendered, pre-determined, and fundamentally passive. You can navigate a game world, but you cannot change its physics. You can watch a video, but the story ends where the file stops. We have spent decades building increasingly high-fidelity cages—beautiful, but rigid.
Today, we are thrilled to announce our investment in Odyssey.
Odyssey is not just building another AI video generator. They are a visionary AI lab building the engine for Generative Reality: a world where digital environments are not fetched from a server, but imagined in real-time, frame-by-frame.
Beyond Video: The Shift to Causal World Models
The current AI video landscape is largely dominated by bidirectional models—systems that look at a beginning and an end to "fill in" the middle. While the results are visually stunning, they are inherently "closed." They are clips, not worlds. They cannot scale to infinity because they require a predetermined "ending" before the first frame is even rendered.
Odyssey has taken a fundamentally different architectural path. By utilizing causal architecture, Odyssey generates each frame based solely on the historical state of the world.
The result is a Streaming World. At 20 frames per second, the latency is low enough that the user is no longer a spectator; they are a participant. This introduces a new paradigm: Steerable Media. You can intervene mid-stream—change the lighting, introduce a new character, or alter the laws of physics—and the world adapts instantly while maintaining Object Permanence.
Solving the "Memory" Problem: Object Permanence vs. Hallucination Drift
Most generative models today have the memory of, well, a goldfish. In traditional AI video can experience hallucination drift. A character might walk behind a tree and emerge as a different person, or a house might vanish when the camera pans away. The model loses the "thread" of reality because it lacks a persistent internal representation of the world.
Odyssey is an architect, not just an artist. Because it is built as a World Model, it maintains an internal "state" of the environment. If a character walks into another room, Odyssey remembers they are there even when they are out of view. This long-horizon coherence is the holy grail of simulation; it ensures that the "common sense" of the world remains intact over indefinite periods.
From Autonomous Vehicles to Virtual Universes
The hardest problems in AI are found at the intersection of the digital and the physical. To build a world model that doesn't collapse into nonsense, you need more than just scale; you need an obsession with state, reasoning, and continuity.
This is why Oliver Cameron and Jeff Hawke are the ideal founders for this mission.
Oliver and Jeff have spent many years solving the "reality gap." starting with the AV world. They are now applying that same rigor—tracking hidden states and operating under extreme latency constraints—to the creative world. They aren't just generating pixels; they are teaching computers the physics of reality.
The New Architecture of Play
We believe Odyssey is the starting gun for a new category of software.
Gaming moves from open worlds to Infinite Worlds that evolve based on a player's unique psychology.
Education evolves from static videos to Interactive Simulations where the environment reacts to a student's input in real-time.
The Creator Economy shifts from consuming content to inhabiting shared, hallucinated spaces that persist and evolve.
We are proud to back Oliver, Jeff, and the entire Odyssey team as they build the foundation for a future where the only limit to reality is our imagination.
