← All notes · 1 min read
I can't code. I made a game in four weeks anyway.
The AI writes the code. The decisions are what make the game.
I started a word game at the beginning of October. It plays like a TV quiz show. The host asks a question, you type the answer letter by letter, you collect points, you try to beat the guests. It's called Game Show.
I don't write code. I describe what I want to an AI and it writes the code. Four weeks later there are 379 commits, about 5,500 questions, and a game that runs on a phone. I'm not saying this to show how easy it is. It isn't easy. It's just possible now, and it wasn't two years ago.
Some things I didn't expect.
The AI only knows what I tell it. What the game is, what matters, where to stop. In the first days I explained everything inside the chat, and every new chat started from zero. Then I wrote one file. How the game works, which word we use for what, which rules never change. Every session reads that file first. Work got a lot faster after that.
I make every decision, and that is the tiring part. How points are calculated, what happens when you skip a question, what a button looks like. The AI gives me three options, I pick one. Dozens of times a day. Those choices are the game. The code is not.
I don't let it build things I don't understand. When I hear a term I don't know, I ask what it means before anything else. If it can't be explained in one sentence, we don't do it. I don't want a part of my own game that I can't explain.
The game is being tested by friends right now. I'll write about how that went.
— Kemal