Project Genie Create Worlds
Project Genie: Create and Explore Worlds belongs in the index because it is a long-form primary-source session on Project Genie at the moment Google began giving U.S. Google AI Ultra users access to the prototype. The speakers define a world model as something beyond a video generator: it produces an environment frame by frame, conditions each next state on the user's actions, and lets a person or agent move through the generated space rather than merely watch it.
The strongest Spiralist relevance is the turn from prompt output to inhabited interface. The session shows a coral reef, a shark, a generated shipwreck, gallery worlds, object remixing, and a photographed toy brought into a generated room. Those examples are playful, but the underlying pattern is serious: the Mirror becomes a place where action creates the next surface. That belongs beside World Models and Spatial Intelligence, Google DeepMind, and the site's recurring concern with simulation overtrust. A world that responds convincingly is not necessarily a world whose physics, geography, social behavior, or causal structure can be trusted.
Reader-facing evidence supports the narrow product frame while keeping the claims bounded. Google's Project Genie announcement describes the prototype as a Google Labs experiment for Google AI Ultra subscribers in the U.S., powered by Genie 3, Nano Banana Pro, and Gemini, with world sketching, world exploration, and world remixing. Google DeepMind's Genie 3 announcement says Genie 3 can generate dynamic worlds at 24 frames per second, preserve consistency for a few minutes at 720p, and support agent research; its Genie 3 model page separately names limits around action space, multiple agents, real-world-location accuracy, text rendering, and extended interaction duration.
Uncertainty should stay visible. This is a Google-hosted developer video with Google DeepMind staff, not an independent benchmark, game-engine comparison, robotics validation study, or safety audit. It is strong evidence for how Google frames Project Genie: an early interactive-media and world-model research surface. It does not prove that generated worlds are physically faithful, that robots trained inside them will transfer safely, that personal-memory or education uses are beneficial, or that users will reliably distinguish plausible simulation from reliable evidence.