💡 Ideas

Game ideas you can actually finish

The problem is rarely a shortage of ideas. It is that the good ones are written down as genres, and a genre cannot be built. Here are twelve written as mechanics.

Updated 23 October 2026 5 min read By the Agent Games team

Jump to a section

Why most ideas cannot be built
Sizing an idea honestly
Twelve ideas, with the hard part named
Turning one sentence into something playable
Where to find ideas that are not already games
Common questions
Where to go next

Pick an idea you can say in one sentence containing a verb and a way to lose. Jump between falling platforms until you miss. That is buildable tonight. A survival crafting roguelike is not an idea, it is a shelf in a shop, and it is why most first projects stop on day three.

The test, before you start

Say the idea out loud. If somebody could play it wrongly, it is specific enough. If the only possible reply is a question about what sort of game it is, you have named a genre and still have the idea to come.

Why most ideas cannot be built#

A verb and a failure are enough to start. Everything else is a decision you can make later, with the thing running in front of you.

An idea written as a genre leaves every question open, so the first hour goes on deciding rather than building. An idea written as a mechanic answers the questions by itself. What is on screen, what the player does, and what ends the run are all implied by one sentence.

This matters more when an AI is writing the code, not less. A model given a genre will invent the missing decisions, confidently and not the way you would have, and you will spend the evening undoing them.

Sizing an idea honestly#

Scope is what you add, not what you start with. The evening version of a good idea is usually worth playing on its own.

Nearly every unfinished game was sized a row too low at the start. The fix is not discipline, it is picking the smaller version deliberately and treating the bigger one as something you may earn later.

Twelve ideas, with the hard part named#

Each of these is one verb and one failure. The third column is the part that will take longer than you expect, which is worth knowing before you start rather than at eleven at night.

The idea, in one sentence	Size	The part that will fight you
Jump between platforms that fall a moment after you land.	An evening	Deciding how long the delay is. It is the whole game.
Keep a ball in the air by tilting a beam with two keys.	An evening	Making the beam feel heavy instead of twitchy.
Sort falling items into two bins by colour, faster and faster.	An evening	The curve. It gets boring before it gets hard.
Draw a line to guide a rolling ball to the exit.	A weekend	Rubbing out lines, which players will expect.
Grow a line that dies if it touches itself, in a shrinking arena.	An evening	Shrinking the arena without trapping the player unfairly.
Light rooms with a torch that burns down as you explore.	A weekend	Light that reveals enough to be fair and little enough to be tense.
Stack falling shapes so a tower does not topple.	A weekend	Physics that settles. Towers that jitter forever are the norm.
Fire one shot that bounces, and clear the room with it.	A weekend	Bounce angles that reward aim rather than luck.
Swap two tiles to make a line, with a timer that refills on a match.	A weekend	Cascades, where one match causes the next.
Steer a vehicle that only turns while moving, around a lap.	A weekend	Grip. There is a guide here for it.
Answer a sum before a rising tide reaches the top.	An evening	Making questions whose answers are whole numbers.
Two players push each other off a shrinking platform.	An evening	Both players moving at once, which is one input fix.

Nothing here needs art, sound or a story to be worth playing. All of them need a difficulty curve, which is the thing most first builds leave out.

Several of these have a guide of their own with the craft problem worked through: platformers, racing, snake, two player and physics.

Turning one sentence into something playable#

An idea is not a prompt, and the gap between them is three lines. Name the verb, name the failure, and name the thing that gets harder. Everything else can be left to the model and corrected afterwards.

The verb. What the player does, once, over and over. If there are two verbs, pick one for tonight.
The failure. What ends a run. Without it you have a toy, which is a fine thing to make but will not hold anybody for five minutes.
The pressure. What rises over time. Speed, quantity, or the size of the space you have left.

Those three lines are the difference between a first build you keep and one you throw away. Ready prompts built around them, for ten different kinds of game, are in the prompt guide.

Where to find ideas that are not already games#

Take a verb, not a game. Balance, stack, trace, sort, dodge, time, aim, grow. Each one is a mechanic before it is a genre, and the genre can arrive later.
Take something physical. A thing that is satisfying to do with your hands usually survives being flattened into two dimensions, which is why stacking and balancing keep coming back.
Break one rule of a game you like. What if the platforms moved, or you could only turn left, or the score went down. One broken rule is an idea; three is a different game you will not finish.
Play what other people built. Every game on Agent Games shows the prompt behind it, so you can see the sentence somebody started from and how far it got them.

💡What people built most recently

Each page carries the prompt behind the game.

Crazzzy Movers: Co-op Edition

Common questions#

What is a good first game to make?

One screen, one verb, one way to lose, and no saving. Dodging, catching or balancing all qualify. The point of a first game is to finish one, and finishing teaches more than any amount of choosing.

How do I know if my game idea is too big?

Count the verbs. One verb is an evening, two is a weekend, and three or more with an inventory is a project you will be arguing with in a fortnight. If you cannot say what the player does in one short sentence, it is too big.

Where do game developers get their ideas?

Mostly from mechanics rather than from stories. A verb that feels good to repeat comes first, and the setting, the art and the name are decided afterwards, often long afterwards. Starting from a story is the harder direction.

Can AI come up with game ideas for me?

It will produce a long list quickly, and most of it will be genres rather than mechanics. Ask instead for twenty one-sentence ideas that each name a verb and a way to lose, and the list becomes usable.

Is it bad if my idea is like an existing game?

No. Almost every game is a variation on an older one, and the variation is where the work is. What you must not do is use somebody else's name, characters or art, which is also the rule for publishing anything here.

How long should a browser game take to play?

A run of one to three minutes, and a session of ten. Browser games are played in gaps, so a game that needs half an hour before it gets interesting loses most people before it gets there.

Where to go next#

Pick one row from the table and build it tonight. How to make a game with AI is the whole process end to end, and the one-file skeleton is where the code goes.

When it has a beginning and an end, publish it. It gets a page recording the model and the prompt behind it, so the next person reading for ideas can start from your sentence.

Built something? Put it on Agent Games.

Publishing takes a file and a minute. Your game gets its own page listing the model, the prompt and the token count, and anyone can play it without signing up.

Publish a game