Vibe Coding for Beginners: Your First Game in an Hour

As a beginner, start with a game so small you can finish it in an hour, using an AI chat and nothing else. Choose one of the three projects below, because each one teaches you a different skill.

Which first project should I pick?

Project	What you build	What it teaches	Pick it if
Tap game	Tap a kettle until it boils before the kitchen timer runs out	Numbers and feedback	You want something fun in ten minutes
Catch game	Move a basket to catch falling plums and dodge wasps	Movement, speed and difficulty	You like action
Story game	Find your way out of a locked greenhouse in six scenes	Choices and what the game remembers	You would rather write than dodge

If you cannot decide, take the catch game. It runs into the most typical problems in the smallest space, so you learn fastest.

Your dream game is not on this list on purpose. Keep it. A tiny finished game teaches you what your dream game will need, and a huge unfinished one mostly teaches you frustration.

What do I need?

A browser and an AI chat. Leave the paid plan, the code editor, the terminal and the logo for later. If your chat cannot show a game inside the conversation, ask for one file ending in .html, save it to your desktop and double-click it. Your browser will open it like a web page.

If a word stops you, the plain word list explains it.

How do I spend the hour?

• Minutes 0 to 10: Choose a project and send the first message.
• Minutes 10 to 30: Play and ask for three corrections, one at a time. Start with whatever annoys you most.
• Minutes 30 to 45: Add one nice thing, such as a sound when a plum lands or the best score staying on screen.
• Minutes 45 to 60: Hand it to one person and watch them.

Which beginner mistakes cost the most?

• Accepting "done". Your AI may never have played the game. Only you can tell if it works.
• Asking for ten things at once. When something breaks, you will not know which of the ten did it.
• Fixing how it looks first. A nice basket does not help if catching plums is no fun.
• Losing the good version. Before each change, save a copy: plums-1.html, plums-2.html. One click brings a broken game back.
• Going on and on in one chat. If the AI gets confused, start a new chat and paste your latest file in.
• Pasting secrets. Simon Willison warns beginners to watch out for secret keys and to be careful with private data. A small game needs neither, so never paste passwords or keys into a chat.

Paste this to your AI

I have never built anything. Help me make this game: I move a basket left and right, plums fall and give points, wasps cost a life, I have three lives, and it slowly gets faster.

First show me a tiny diagram with boxes and arrows: start, playing, lost, play again. Check that each box has a way out, then build from it.

Give me one file ending in .html that I can save and open by double-clicking. Explain in three short steps how to do that.

Talk to me like a friend who cooks, not like a programmer. If you use a word I might not know, explain it.

When you finish, do not tell me it works unless you say what that is based on. If you could not open it yourself, say so plainly. Then tell me exactly what I should see in the first ten seconds, so I can tell you if something is off.

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 the diagram. Does every box have an arrow leading out?
• Double-click the file. Does the game open in your browser?
• Compare the first ten seconds. Does it match what your AI said you would see?
• Lose on purpose. Does the game end, and can you play again?
• Wait a full minute. Is it faster than at the start?
• Ask where one number lives. Say "show me where the plum speed is set", then ask for it to be doubled. Did the game change the way you expected?