Genie 3, developed by Google DeepMind, allows users to create interactive 3D environments using brief text prompts. It enables exploration and manipulation of these virtual scenes for a limited time, maintaining stability and coherence for over a minute, unlike its predecessors, Genie 1 and 2. Genie 3 can respond instantly to new commands, allowing real-time changes without reloading. Currently, it is in research preview and accessible only to select academics and creators. The system has limitations, including basic interaction mechanics and challenges with multiple agents and accurate real-world replicas.