After months of waiting, I finally got access to Google DeepMind’s Genie (Genie 3), the real-time world simulator that is redefining Generative AI. This isn’t just video generation anymore—this is a playable, interactive environment generated on the fly.
In today’s video, I take you hands-on with the public release to stress test the model. We go from “Google Street View” prompts to highly stylized comic book pages that come to life, test the limits of object permanence (Elden Ring style), and see if we can break the physics engine.
I also sat down for an exclusive interview with the DeepMind team to discuss the tech stack, why certain features were cut from the launch, and their roadmap to simulating reality with Genie 4.