Google has launched Project Genie, a new AI tool allowing users to create and explore fully interactive virtual worlds. The project, developed by Google DeepMind, is currently available exclusively to subscribers of Google AI Ultra in the United States, costing $249.99 per month.

How Project Genie Works

Project Genie isn’t just a static environment generator; it lets users build worlds and embody a character within them. Users can use text prompts to design custom levels, including simple mini-games playable with standard keyboard controls. One reporter at The Verge demonstrated this by quickly recreating basic versions of Zelda and Super Mario. The system generates worlds at 720p resolution and 24 frames per second.

This is significant because traditional video game development requires substantial time and resources. Project Genie delivers detailed, interactive environments with minimal effort, though Google emphasizes it remains primarily a research project at this stage.

The Underlying Tech: World Models and AGI

Project Genie is powered by Genie 3, Google’s experimental “world model.” A world model is an AI capable of generating virtual environments from various inputs—text, images, and more—while also simulating realistic physics. Google first hinted at Genie 3 in August 2025, describing it as a crucial step toward artificial general intelligence (AGI).

“Genie 3 represents a major leap in capabilities… enabling AI agents capable of reasoning, problem-solving, and real-world actions.”

AGI is the hypothetical point where AI can perform most human tasks as effectively or better, with minimal oversight. Developing effective world models is seen as vital for achieving AGI: AI must understand and interact with environments before it can function autonomously.

Beyond Gaming: Real-World Applications

The immediate use case is entertainment, but the implications extend far beyond. The AI sector has seen massive investment in world model technology; WorldLabs recently secured $230 million, while Luma AI raised $900 million. Potential applications include:

  • Autonomous Vehicle Testing : Safe, controlled environments for testing self-driving cars.
  • Education : Immersive simulations for training or historical recreations.
  • Game Development : Accelerating prototyping and level design.

Currently, Project Genie is a closed beta for Google AI Ultra subscribers in the US. However, the underlying technology points towards a future where AI-generated virtual environments become commonplace.