From prompt to playable
Users describe a game to a chatbot; Playground builds it in the browser. They can keep iterating with new prompts, upload photos, video or models as references, or let AI generate the artwork and design. The conversational process is simple enough for someone with little experience in game development or programming.
The first result was Pork Drop, prompted as “Tetris, but with cute pigs instead of blocks.” A follow-up asked for the blocks themselves to be pigs, twisted into Tetris-piece shapes. The result was tidy and playable, though some pigs looked like they had frozen in a T-pose. Playground suggested the name Piggy Stacker; I kept Pork Drop.
A colleague, Jay Deirit, suggested a “gay dinner party.” Playground turned that into Dinner Darling, a two-dimensional game with Bitmoji-like characters. You play Julian, an exhausted host trying to serve picky guests bruschetta and espresso martinis before their patience runs out. I failed quickly, still fumbling with the controls when the game declared: “DINNER PARTY DISASTER.”
The game never mentions the characters’ sexuality. After my loss, Deirit shouted that I was bad at it because I was straight. “Not bad, Playground,” I thought. Another colleague, Rhys Rogers, was less impressed: he said the game was unrealistic because it offered no vegan or dairy-free food.
Neon Don and the iteration loop
The most involved project became Neon Don, a top-down arcade shooter with synthwave music. I first asked for a cyberpunk rhythm game starring a “heavily divorced” former cop, then asked Playground to remake it as a shooter like Hotline Miami. By publication, seven different players had tried it—enough, in my view, to call it a masterpiece.
Playground supplied some of the enemy description itself: “big, broad-shouldered, with prominent beer bellies stretching tight tactical shirts and cargo pants,” plus “a special tough-guy walk.” I kept feeding it bugs and feature requests. It fixed movement problems, including a character getting stuck, but sometimes introduced new ones.
I also asked for sarcastic lines the protagonist could deliver in battle, including “Keep shooting! Brenda took everything else from me!” and “I’ve taken worse hits from my ex-wife’s lawyer!” The result was amusing. It was not the same as knowing how to specify a feature well enough to make a game more interesting or usable.
My own prompts tended to be broad: “Make Don taller,” or “Make him even more divorced.” Some of the stronger games on the platform are noticeably better than mine. The distinction may not be the ability to generate a game, but the ability to guide it.
The ceiling is visible
Playground limits how much users can build and revise: each new game and each change uses up an allowance. I told my new fans I would update my games on October 14, when my allowance resets. The backflip button I wanted for Neon Don was labeled “Doge flip”; changing it will have to wait. I also want to tune the difficulty and add weapons, but I am not paying for more prompts.
The more interesting question is what happens when the novelty wears off. Across genres, games tend to share menus, music, combat options and visual choices. Playground makes it easy to get something running, but the ease of generation also makes the results feel interchangeable.
I think the tool is a compelling way to experiment with how games are made. But its usage limits make iteration scarce, while its repeated design choices make the output feel less like a world of games than a small set of templates. That is a thin foundation for turning a quick prompt into a distinctive game.
Daily AI news
Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.
Only what matters — every day
Follow on X