I Made Terrible Games With Google’s AI Playground

Google has launched Playground inside Google Labs, its home for AI experiments that have not graduated into full products. It is vibe coding for video games: describe what you want in plain language, and it builds a working browser game you can keep refining by chat, with uploaded photos, videos, or models as optional references. Meta announced a similar service, Horizon Create, in September.

A WIRED reviewer with little development or coding experience built three playable, stable games in a single day, from a pig-themed Tetris clone to a top-down shooter starring a very divorced ex-cop. Anyone with a Google account can sign in and play what others make. The barrier to making a game just dropped to roughly zero.

We like what this unlocks. The reviewer says Playground made him curious about the actual craft of game-making, and that is the best case for tools like this: a fast, low-stakes loop where an idea turns into something playable in minutes. For kids, hobbyists, and teachers, that loop is a real gift. It is also a decent prototyping sketchpad for people who already know what they are doing.

But the review is honest about the ceiling. Iteration is conversational, and conversation is imprecise. Fixes sometimes introduce new problems. The model invents details nobody asked for; a requested backflip button came back under a different name, and fixing it had to wait for the usage allowance to refresh. Google throttles usage, every build and tweak spends from an allowance, and the reviewer declined to pay for more. That pricing model shapes what gets made: quick drafts, not finished work.

The bigger issue is the one the reviewer lands on at the end. Scroll the public games and the menus, music, combat options, and style choices start to look identical across genres. That is what happens when one model fills in every blank the user leaves open. Multiply that across a platform where anyone can publish, and you get a flood of competent, interchangeable games. The reviewer's own verdict was slop all the way down.

I'm watching where this leads if it scales. For players, discovery gets harder as volume rises. For working developers and artists, the risk is less that Playground replaces them and more that it floods storefronts and feeds with look-alike output, making distinct work harder to find. The fix is not to block the tools. It is to make the human's choices matter more: real control over assets and parameters, clear labeling of AI-generated games, and curation that rewards originality over volume.

Our read: Playground is a good toy and a promising teacher. It is not yet a creative tool, and Google should not pretend otherwise.

Google Labs has previously hosted experiments like AI music production before they become standalone products. Playground's visuals are AI-generated by default unless users upload their own reference assets, and its games run in the browser.