Meta’s Pocket is built around a simple promise: describe an interactive idea, wait briefly, and get something playable on a phone. For people who are curious about AI game creation but do not want to deal with CLIs, IDEs, terminal work, or setup, that changes the starting point.
The tradeoff is just as important. Pocket makes creation feel immediate, but the finished gizmos live inside Meta’s social environment, where sharing is easy and taking the work elsewhere is not.
Prompting replaces the usual coding setup
Pocket launched in the US last Friday, August 21, after Meta acquired the staff behind the now-defunct vibe-coding app Gizmo in March. The app lets users make fully functional interactive “gizmos” from text prompts, with no visible code and no option to inspect the code behind the result.
That matters because AI-assisted coding and “vibe coding” have already changed how software projects can be started. But many of those workflows still ask beginners to install tools, use a terminal, manage project files, and understand at least some developer environment basics.
Pocket removes that layer. The creation flow begins with a large text box labeled “Describe an idea…” and turns a written request into a mobile experience. Once made, gizmos can be shared in an endless social feed with likes, comments, and reposts.
Fast prototypes can become real game experiments
The source test began with a simple game concept: tap and hold the phone screen to spin up a sling, then release to fire a tangential shot at approaching enemies. The baseline prompt was: “An overhead shot of a man in the center of the screen. When you tap and hold he rotates a sling around him in a circle. When you lift your finger the sling sends out a ball perpendicular to the sling. The longer you hold the faster the ball comes out.”
After nearly two minutes, Pocket produced a functional version that matched the broad idea. It also added elements that were not requested, including explodable hay bale targets, a score counter, and an on-screen power meter.
That first result became the foundation for more than 100 prompts. The project grew from a basic sling-shot control scheme into something closer to a full game, with robots entering from the screen edge, multiple robot types, items, an electric fence around the player, changes to scoring, sound effects, sling speed, and enemy patterns.
A second project followed: a simple arcade-style juggling game with colorful balls kept in the air by thumb-controlled paddles. More prompts added items, a complex score-multiplier system, and end-of-game graphs for tracking progress.
The creative loop is the main attraction
Pocket’s strongest idea is the loop between asking and testing. A creator can think of a feature, describe it, and see a working version about a minute later. For rough game design, that compresses the distance between an idea and a playable test.
The results were often close to what was requested. Some prompts needed follow-up clarification, and some ideas worked better in theory than they did in play. Only a handful of prompts were described as being ignored or as breaking the game entirely, and a single tap could restore the last working version.
The app also includes an “Ideas” tab that suggests possible feature additions. Those suggestions were inconsistent. Some did not fit the game, and some repeated existing features in a weaker form. Others added useful presentation effects or gameplay variations that had not been considered yet.
That makes Pocket feel less like a blank coding environment and more like a collaborative prototyping surface. The user still chooses direction, but the app can generate both implementation and suggestions.
The missing code creates real limits
The same abstraction that makes Pocket approachable also limits precision. A small layout change, such as shifting a UI element a few pixels, still requires a prompt and another rebuild. In a hand-coded project, that kind of adjustment would usually be direct and quick.
Testing is another weak point. Without code access, memory traces, or execution stops, debugging depends on watching behavior and using on-screen displays. That can be enough for a toy project, but it makes hidden bugs and edge cases harder to find.
The output also tends to have a generic look and sound. Pocket often defaults to emojis or basic shapes for visual elements, while WebAudio effects lean toward simple beeps and boops. Users can add recorded sounds or ask Pocket to generate graphics from AI prompts, but the baseline style can still be hard to escape.
Meta gains the easiest sharing path
Pocket’s social design is central to the product. Gizmos are not just generated; they are meant to circulate inside a TikTok-style feed. That creates a direct path from creation to audience, with engagement features already built around the work.
But this also creates the clearest concern. The more useful Pocket becomes as an AI coding tool, the more valuable the resulting work may feel to the person who made it. Yet the gizmos described in the source are effectively trapped in Meta’s environment, with no real path to turn them into independent projects.
For casual creators, that may be acceptable. Pocket can turn a phone, a prompt, and a game idea into something playable very quickly. For anyone hoping to graduate a prototype into a separate app or serious project, the lack of code access and export flexibility is not a small detail. It defines what Pocket is: a powerful social creation tool, but not a full replacement for owning the software you make.