Vibe Coding a Game in 7 Days: A Day-by-Day Plan

Vibe coding a game in 7 days works if you keep it small, decide on day 1 what you will cut, and spend the last two days finishing instead of adding. Here is the plan, worked through on an invented game where you keep a beehive alive through one summer.

Allow about two hours a day. Each day ends with a clear "done when", and if it is not done, the next day starts there.

Day 1: size it and map it

Write the game in one sentence: "Keep your hive fed until the first frost." Then have your AI draw a full run as a diagram: how a summer starts, what can happen, how it ends, how you play again. Look for dead ends. If the hive can starve and the game just stops, you have found a missing box before writing a line.

Sort the boxes into must-have, nice-to-have and cut-first.

Done when: the map has no dead ends, and your cut list exists.

Day 2: the ugly core

Bees fly out, honey comes back, a number goes up. Squares are fine. Do not touch colours today.

Done when: you can play one week of summer.

Day 3: the one hard thing

Every game has one thing that makes it a game. Here it is choosing how many bees stay to guard the hive while the others gather. Spend the whole day on that.

Done when: you have made a choice in the game that you had to think about. If nothing is interesting by tonight, change the game, not the schedule.

Day 4: an ending

Add losing, winning and starting again. This is less fun than day 3 and more important.

Done when: you can play from the first day of summer to the frost and start again without reloading.

Day 5: stop adding

At noon, feature work ends. After that, you only improve what is already there: colours, sounds, a title, a screen that explains the controls.

Done when: it looks as if someone meant it to look like that.

Day 6: strangers

Three people play while you say nothing. Write down where they stop. Fix only what stopped them, however much the rest bothers you.

Done when: a new player can finish without asking you anything.

Day 7: publish

Put it online, share one link, and stop. Your next idea goes in a notebook. Getting a game into a browser explains the steps.

Done when: someone you do not know could play it.

What do I do when I fall behind?

You will. Here is the rule: days do not move, the cut list does. If day 3 ends with no fun in it, take the top item off the cut list and carry on. For the bees, the cuts might go like this: first the names for each bee, then the weather, then the shop.

Never cut losing, starting again or the map. A small game with an ending counts as finished, and a big one without an ending does not.

If you have a weekend and not a week, the game jam plan works in hours. And when a mechanic takes all day, see if a blueprint already covers it, or browse by topic.

Paste this to your AI

I have seven days for this game: [one sentence]. About two hours a day.

Before we write anything, draw one full run of the game as a Mermaid diagram: how it starts, what can happen in the middle, how it ends and how you play again. Go through it looking for any situation the player cannot leave.

Then sort the boxes into three piles by what the game cannot live without: must-have, nice-to-have and cut-first.

For each box, say why it is in that pile and how long you guess it will take, marked as a guess. If you have nothing to base a guess on, say so.

Stop there. I will look at the map before day 2.

Or skip this and connect the MCP: your AI then finds all of this by itself.Connect the MCP

How to check it worked

• Read your cut list out loud on day 1. Could you lose those things without being sad?
• Play a full run on day 4. Can you go from start to end with no tricks?
• Look at the clock on day 5. Did anything new go in after noon?
• Compare one guess with reality. How far off was the AI? Adjust the rest.
• Open your link on someone else's phone on day 7.