Kennsley Staniszewski
Peters, J. (2026, January 29). Google’s AI helped me make bad nintendo knockoffs. The Verge. https://www.theverge.com/news/869726/google-ai-project-genie-3-world-model-hands-on
Google’s Project Genie is a new experimental tool that uses the Genie 3 AI model to generate interactive 3D worlds from text or image prompts. The tool is rolling out to Google AI Ultra subscribers in the US and represents Google DeepMind’s work on AI “world models” that can create virtual interactive spaces. Users can either choose from the pre-designed worlds or create their own by writing prompts that describe environments and characters. Once generated, these worlds run at 720p resolution and 24fps, and users can explore them for 60 seconds using keyboard controls. The AI generates frames in real-time based on user movements rather than creating pre-rendered video. The Verge reporter, Jay Peters, tested the tool and found several limitations. There’s noticeable input lag, worlds sometimes lose consistency (forgetting previous changes or suddenly altering terrain), and the 60-second time limit restricts meaningful exploration. The reporter also discovered that the model, trained on publicly available web data, could initially generate worlds based on copyrighted gaming franchises like Nintendo properties, though Google began blocking these requests. Overall, it is an impressive work-in-progress in the world of AI, but it’s not yet at a level where it can compete with traditionally designed interactive video games.
I thought this article was a fun read because it shows what happens when someone actually gets their hands on new AI tech and just messes around with it. The reporter spent his time making bootleg Nintendo games, which I thought was pretty fun. It’s refreshing to see a real test of the technology instead of just reading about how amazing it’s supposed to be! The videos were really interesting to watch as well! This was yet another new addition to the AI world right now that feels like a pretty different approach and could eventually be useful for things like education or even training robots to navigate spaces! The article also touches on some messy copyright issues which is something I have been wondering about! The AI was trained on public web data and could generate worlds that looked a lot like Mario and Zelda games!! The technology is cool but still pretty rough around the edges, which feels kind of comforting and important to remember when everyone’s talking about how AI is going to change everything overnight and when it feels like there are leaps and bounds made in artificial intelligence everyday!