Write the prompt around a playable rule
Give the prompt a bounded setting and an action the player can repeat. Specify what changes after success, what the player risks, and how the screen communicates a good move. The tower example defines a loop without demanding a complicated campaign. Keep optional ideas in a separate note so the first generated version can focus on the basic drop-and-trim rhythm.
Original concept prompt: “Make a one-button tower game. A slab slides over the current tower; tap to drop it. Any overhang is trimmed, and the remaining width supports the next slab. Show a best height, make perfect landings visibly rewarding, and end the run when no tower remains.” This is a recipe for an idea, not a claim that this exact game is already playable.
Play, observe and refine one thing at a time
Play enough attempts to see both a clean landing and a badly trimmed one. Check that the slab movement gives a fair cue, the remaining platform is visible, and the run ends for a reason you understand. If misses feel arbitrary, request slower movement or a stronger landing cue. If every drop is easy, tune speed or slab width, not several variables at once.
Change the prompt according to a specific moment from play. Ask for “a brief pause before the slab reverses direction” if the cue is too subtle, rather than “make controls nicer.” Generation may take time; treat each version as something to inspect and test. Creating and remixing games requires Creator, so review pricing for current plan information.
A first version need not contain every theme, level or scoring idea. If the main action is enjoyable but the run lacks variety, ask for one modest change such as a speed increase after several drops. If the main action itself is confusing, reduce the scope and clarify the rule before adding content.
Playable reference for this prompt: Stack. Tap to drop each sliding slab onto the tower; any overhang is sliced off, so perfect drops keep it wide. Use this example to compare the proposed rules with a game you can try now.
Questions about make a game with AI
How small should my first game idea be?
Start with a complete loop that fits on one screen. The tower example needs a moving slab, a drop action, an overlap rule and a restart.
When should I add another feature?
Wait until you can start, understand, finish and restart the basic game. Then add one feature and check whether it improves the decision you wanted players to make.