Google has launched Project Genie, an experimental research prototype that gives select users the ability to create, explore, and remix interactive, AI-generated worlds using its advanced Genie 3 world model. Available today to Google AI Ultra subscribers in the U.S. (18+), the prototype marks a significant step in the development of general-purpose AI systems capable of simulating dynamic, real-world environments. Unlike static 3D experiences, Project Genie generates environments in real time, adapting as users move and interact, with physics and object behaviors simulated on the fly. Genie 3 is designed to be a foundational step toward artificial general intelligence (AGI), moving beyond narrow AI agents trained for specific tasks like chess or Go. Instead, it aims to model the complexity and diversity of the real world by predicting how environments evolve and how actions influence them. This capability enables simulations of anything from historical settings and fictional landscapes to robotics training scenarios and animated sequences. The system’s breakthrough consistency allows it to maintain coherent, long-term world states even during extended interactions. The prototype is powered by Genie 3, Nano Banana Pro, and Gemini, and operates as a web-based app. It centers on three core capabilities: generating new worlds from text prompts, exploring those worlds in real time, and remixing existing environments with new elements. Users can build worlds in 60-second increments, shaping terrain, adding characters, and introducing interactive events. The experience is dynamic—objects respond to gravity, collisions, and user actions, creating a sense of immersion and realism not found in traditional 3D renderings. Google emphasizes that Project Genie is an early-stage research prototype, part of its broader commitment to responsible AI development. While the model shows promise, it is not yet fully mature. Some features announced in August, such as promptable events that dynamically alter the world during exploration, are not yet available. The company also acknowledges limitations in object consistency, long-term memory, and realism, and continues to refine the model through feedback from trusted testers across industries. Access is currently limited to Google AI Ultra subscribers in the U.S., reflecting the prototype’s experimental nature and the computational demands of real-time world simulation. Google plans to expand availability to more regions in the future, with the long-term goal of making these immersive AI experiences accessible to a wider audience. This initiative builds on Google DeepMind’s prior work in AI agents and world modeling, and represents a bold leap in generative AI’s potential. By enabling users to co-create living, evolving environments, Project Genie could influence fields ranging from education and entertainment to design and scientific simulation. It also raises important questions about creativity, ownership, and the ethical use of AI-generated content. As users begin to experiment with the prototype, Google will gather insights on how people interact with AI-generated worlds, helping shape the future of interactive media and general AI. While still in early development, Project Genie offers a glimpse into a future where AI doesn’t just respond to prompts—but builds and sustains entire worlds in real time, transforming how we imagine, explore, and create.