Summary: Google opened Project Genie to US Google AI Ultra subscribers aged 18 or older on January 29, 2026, as the first public product built on DeepMind’s Genie 3 world model. It generates interactive 3D worlds from prompts and images but caps each generation at 60 seconds and remains a research prototype. The next day, Reuters reported gaming stocks falling, with Unity down nearly 24% intraday.
January 29, 2026 is the day Google put Project Genie into the hands of Google AI Ultra subscribers in the United States, provided they are 18 or older. Google Labs built it as an experimental research prototype, and it doubles as the first product the public has ever seen on top of Genie 3, DeepMind’s general-purpose world model. You supply a text prompt along with images that were generated or uploaded, and the system hands back a 3D world you can move through and interact with, assembled on the fly.
The jolt arrived a day later. In Friday afternoon trading on January 30, videogame stocks slid across the board: Reuters reported Unity crashing nearly 24% intraday to about $29.75, its worst one-day drop since 2022, while Roblox and Take-Two Interactive — the publisher behind Grand Theft Auto — slid alongside it, a selloff The Verge documented company by company. How does a research prototype capped at 60 seconds per generation push an entire industry to reprice? The sections below unpack where Genie 3’s abilities end and what the shift means for the business.
Genie 3: A Research Preview Turns Into a Shipped Product
A world model does what the name promises: it simulates the dynamics of an environment, predicting where a scene is heading and how your actions bend that course. DeepMind brought Genie 3 out for its first preview in August 2025, cast it as a general-purpose world model, and hooked it straight to the AGI agenda — a general intelligence has to handle the variety of the real world, not win only inside one closed environment like chess or Go. Project Genie is the first place that research meets the public: a web application that combines Genie 3 with Nano Banana Pro — which handles image generation — and Gemini. The United States gets first access, Google AI Ultra for Business accounts are pointedly left out, and further territories are on the plan.
Sketch, Explore, Remix: What You Can Actually Do
Three capabilities carry the whole experience. World sketching: you put together a character and an environment out of text and images, Nano Banana Pro takes care of previewing and fine-tuning the world image, and you choose between a first-person or third-person view plus movement styles like walking, flying, or driving. World exploration: the model “generates the path ahead in real time” — you are not standing inside a pre-rendered static 3D scene but inside a world that grows segment by segment as you act, with a camera you can adjust freely. World remixing: prompts change existing worlds, a curated gallery and a randomizer are on hand for inspiration, and your sessions can be downloaded as video recordings.
What the Market Did: Gaming Stocks Sank Together
For all the product’s novelty, the trading on January 30 may be the more instructive record. Videogame shares came under heavy afternoon selling once Alphabet’s announcement landed, Reuters reported; The Verge counted Take-Two, Roblox, and Unity all finishing Friday in the red. The reasoning behind the selloff is not hard to assemble: the moats that an engine maker (Unity) and a user-generated content platform (Roblox) defend sit precisely where the “one prompt, one playable world” narrative strikes. Whether the market was pricing real disruption or raw panic is still being fought over — among the busiest Hacker News threads was one debating whether GTA VI’s expected revenue could relieve the pressure on Take-Two at all.
The 60-Second Ceiling, Plus a Candid List of Shortcomings
Google’s own account of the weaknesses is refreshingly frank: worlds can fall short of realism, wander off your prompt or uploaded images, or break real-world physics; character control can be sloppy and laggy; every generation stops at 60 seconds; and “promptable events,” demonstrated in the August 2025 Genie 3 announcement, did not make it into the product. In other words, Project Genie as it stands is a research prototype, not a substitute for a game engine. The development worth weighing is a different one: “interactive world generation” has moved from the pages of a paper to a web page that any AI Ultra subscriber can open — and the official line is that more territories and broader access are coming over time.
Three Working Judgments for Developers
Judgments worth taking away. One: the wall between world models and video generation keeps coming down — what Genie 3 outputs is, at its core, video you can act within, which makes it a fresh source of material for games, animation, and robotics simulation. Two: the cost structure of content platforms is in for a rewrite — the moment an interactive scene’s marginal cost nears that of a single API call, UGC platforms must argue all over again for the value of their matchmaking and community. Three: keep a 60-second prototype out of your production pipeline — its proper role is validation, a quick way to test level designs, scene prototypes, and interaction hypotheses. For the industry’s broader signals heading into 2026, our opening-of-year outlook has the wider view; on the world-model track, Project Genie is the first checkpoint opened to the general public.