Vibe Coded Games Examples: 6 Ideas You Could Build
The vibe coded games that work best are small, built around one clear idea and light on art. Here are six invented examples of that kind of game, each with the part that would really be hard.
These are ideas we made up, not real releases. We do not list real games here, because from the outside you usually cannot tell how a game was made. To see real ones, itch.io lets players filter for games tagged as made with AI.
Six examples
1. Night shift at a tiny museum. You point a torch at exhibits that only move when nobody is looking. One screen, one action, a lot of tension. The hard part: what each exhibit does while it is unseen, and making that feel fair.
2. The last train announcer. A text game. You decide what to announce on a late platform, and thirty passengers react to what you said and what you left out. The hard part: keeping track of who heard what. Your AI will lose this unless you write it down, which is what a story bible is for.
3. Soup stall in the rain. You mix three ingredients for customers whose moods change with the weather. The hard part: balance. If one soup pleases everybody, the game is over. Expect hours of tuning numbers while you play.
4. The island you draw. You sketch paths with your finger, and a tiny walker follows them to reach a lighthouse. The hard part: what the walker does with a gap, a loop or a line that ends in the sea.
5. Two players, one keyboard. A tug of war with a rope, each pulling with quick taps on their own side. The hard part: fairness when both press at once. Keyboards and timing can favour one side without anyone noticing.
6. The clockmaker's bench. You fit gears so that every hand on the clock moves. Twenty puzzles. The hard part: every level must be solvable. Your AI will claim it is. Have it walk through the steps of a solution for each level, or show you where one gets stuck.
What do these have in common?
| They have | They avoid |
|---|---|
| One screen, or close to it | A huge world |
| One main action | Ten systems that all talk to each other |
| Simple shapes or text that look deliberate | Art that has to match across hundreds of items |
| Play on one device | Online play with strangers |
Online play is the biggest trap. It looks like one feature, and it brings servers, cheating and players who disconnect at the worst moment. Leave it for a later game.
Why not rebuild a game I love?
Because you learn less, and it can get you into trouble. Copying a game's name, characters or art is a different thing from using an idea, and the line is covered in Can you copy game mechanics?
Your tools may also refuse. On 30 October 2026 Game Developer reported that Anthropic confirmed it steers Claude Code away from building recreations of existing games. Work of your own is meant to keep working: original games, genre games, and games from your own designs and assets. A twist of your own avoids the whole question.
Paste this to your AI
I like this idea: [one of the six, or your own]. Help me turn it into something of my own.
Give me three twists that change what the player does, not only how it looks. Do not base them on any game you know. If a twist reminds you of an existing game, say which one and replace it.
When I pick one, draw only the hard part first, as a Mermaid diagram: every situation in it, what leads to the next, and what must always be true. Go through it and show me one path where it could go wrong. If you cannot find one, tell me what you looked at. Build from that diagram only after I agree.
Or skip this and connect the MCP: your AI then finds all of this by itself.Connect the MCP
How to check it worked
- Tell a friend your twist. What does it remind them of? If they name one game right away, change it.
- Look at the diagram. Is the hard part in there, or only the easy parts?
- Try the path your AI flagged. Does the game do something sensible?
- Ask a fresh chat. "Which existing game is this closest to?" Decide if you are happy with that answer.