How to start making a game: the first step is on paper

What is the first step to making a game?

Writing the idea in three sentences, on a sheet of paper: what the player does, how they win and why they would play again. Then cutting until the game fits on one screen and in a short round.

It sounds like very little, and that is on purpose. People who want to make a game tend to start in the wrong place: they download a program, watch videos, open the AI and type "make an adventure game". An hour later they have something on the screen, and no idea what to do with it.

The three sentences solve that before you spend time or money. They become the yardstick for everything that comes after: what goes into the game, what stays out and when it is done.

Why start on paper and not in the software?

Because AI builds what you describe, and fills in whatever you do not describe with the most common choice. Without a clear idea, the game comes out like a thousand others.

The Japanese artist Kazuki Takakura made an action game with AI writing all the code, without knowing how to program. In the account he published in July 2026, he says that letting AI decide what is fun tends to produce generic work, made for average taste. In his game, the direction, the art, the sounds and the final say stayed with him⁶.

A 2026 academic test shows the same thing from another angle. The researchers asked several AI agents for complete games, across 140 tasks. The best one scored 41.46 out of 100. The agents tended to get a recognizable mechanic right, and delivered little content and presentation that did not hang together⁸.

Our take: the part AI does worst is exactly the part that fits on a sheet of paper. Deciding what the game is, and why someone would play it, is still your job, and you can do it today, without installing anything.

What are a game's three sentences?

They are the answers to three questions. Each one in a single sentence, in everyday words.

1. What does the player do? The main action, with a verb: jump, dodge, fit, guess, plant. If the sentence needs an "and also", the game is still too big.
2. How do they win? The goal, or how they score points. The opposite comes with it: how they lose. Without losing, there is no challenge.
3. Why would they play again? The reason to come back: beating their own record, a new level, a daily challenge, a round so short you want to try one more.

The third sentence is the one most people forget, and it is the one that separates a game from a demo. A game people finish and never open again may be done right, but it does not hook anyone.

In the games we made with AI, which the portal covers behind the scenes, the one-line description already carries the first sentence and what gets in the player's way. Frete Extremo, for example, is a delivery truck in the Amazon, with mud that sinks the wheels, fragile cargo, rain and night. None of these games has been published yet.

How do famous games fit in three sentences?

They fit well, and that is why they work as examples. First, a warning: none of these games was made with AI, and their creators already programmed or worked in games. They are here to show the size of an idea that works, not to promise results.

Wordle. Josh Wardle, a software engineer from Wales, made the game for himself and his partner, Palak Shah. After each guess, every letter turns green (right letter in the right spot), yellow (the letter is in the word, but somewhere else) or gray (it is not in the word). On November 1, 2021, 90 people played; by January 2, 2022, there were more than 300,000².

Flappy Bird. The bird moves right on its own and falls; it only rises when the player taps the screen. Hitting a pipe or the ground ends the round¹. Its creator, Dong Nguyen, an artist and programmer from Vietnam, made the game in a few days, according to Wikipedia, and released it for the iPhone in May 2013¹.

Tetris. Alexey Pajitnov, a researcher at a computing center of the Soviet Academy of Sciences, created the game in the mid-1980s. The player moves and rotates falling pieces, and each complete line disappears and scores points³. The story of how it and other games got started is in famous games that started small.

Notice what the three have in common: a single action, a rule you can explain in one sentence and a short round. None of them started with a big map, a long story or lots of characters.

What is the right size for a first game?

The smallest thing that is still a game: one screen, one main action, a round of a few minutes and a single player. Anything beyond that can come later, once the core is fun.

People who have made a game with AI without knowing how to program say the same:

• Takakura recommends starting with rules the size of an 8-bit console game, in a type of game you already like. For him, simple games still have plenty of ideas left to test, and they do not need to be 3D or full of effects⁶.
• Jonathan Bjerk, an accountant in Norway who released a game made with AI on the web, iPhone and Android, advises against starting with a game where several people play together: first, make the round fun for a single player. And cut ruthlessly, because every extra thing has to be maintained and balanced⁷. In his project, sync bugs between players took up all of September and October⁷.

Small does not mean silly, either. The characters in Thomas Was Alone are colored rectangles. Its creator, Mike Bithell, who was already a professional game designer, made the game in his spare time, with the blessing of the company where he worked, and it passed 1 million copies by April 2014⁴.

And small can serve the right audience. Masako Wakamiya, a retired Japanese bank worker who learned to program after 80, with books, tutorials and a remote teacher, made Hinadan, a game for people over 60 in which the player places the dolls of the Japanese Doll Festival on the right steps. By August 2017, when she was 82, the game had about 42,000 downloads⁵. That was before AI, and it is not a money story: it is a story of a small, clear idea made for someone.

What are the most common mistakes in a first game?

The most common is starting too big. The others almost always grow out of it.

• Starting with your dream game. The open world, the long story, the twenty characters. That is the game you want to make someday, not your first one. Save the idea and make a small piece of it first.
• Starting with a multiplayer game. Playing together online multiplies the work and the bugs. The advice from people who have been through it is to leave it for later⁷.
• Picking a type of game that depends on precise timing. Takakura names rhythm and fighting games as hard for a first project, because they demand precision in timing and in contact between characters⁶.
• Letting AI pick the idea. It makes good suggestions, but if it decides alone, the game drifts toward the ordinary⁶.
• Forgetting the third sentence. A game with no reason to come back gets played once.
• Showing the game only at the end. Bjerk advises getting real people to play early and often⁷. That applies even to the three sentences: if the person reading them cannot picture the game, they are not clear yet.

The other side

Three sentences do not make a game, and a well-written idea does not guarantee it will be fun. You only find that out by playing and watching other people play (our take). The famous games on this page are survivors: for each one, there are thousands of small games almost nobody played. Starting small is the most common way to start, not a shortcut to success.

And after the three sentences?

Then come the software, building with the AI, testing, the art and sound, and the store. The stages, with what AI does in each one and how much it costs, are in how to make a game with AI. Before that, the plan below fits in an afternoon.

Your action plan

Check off what you've done. It's saved on this device, so you can pick up where you left off.

1. Grab a sheet of paper and keep the computer off for this step.
2. Write, in one sentence, what the player does, with an action verb.
3. Write how they win or score points, and how they lose.
4. Write why they would play again.
5. Cut everything that does not fit on one screen and in a round of a few minutes, for a single player.
6. Show the three sentences to someone and ask whether they can picture the game.
7. Keep your dream game on a separate list, to make later.