I Tried Making an Idle Browser Game With AI. The Prompt Was the Easy Part.

Idle games look simple from the outside. Click a thing. Number goes up. Buy upgrade. Number goes up faster. Leave the tab open. Come back later and feel oddly proud of a tiny fake economy. That simplicity is exactly why they are interesting to prototype with AI.

A good idle game does not need a huge map, complex combat, or a cinematic intro. It needs one understandable loop that starts working almost immediately. If the player can understand the loop in 10 seconds and feel a decision in 30 seconds, you probably have something worth testing.

I wanted to see how far I could get with one prompt and a browser-first workflow. Not a polished commercial game. Not a full economy sim. Just a playable idle prototype that could answer one question: Does this loop feel worth repeating?

The prototype goal

The goal was not "make an idle game." That is too vague. The goal was: Make a tiny browser idle game where the player runs a weird moon farm, harvests glowing mushrooms, sells them for credits, and buys upgrades that make harvesting faster.

The prompt I used

Here is the kind of prompt I would start with:

Create a simple browser idle game called Moon Mushroom Farm. Core loop:

• The player clicks a mushroom patch to harvest glowing mushrooms.
• Mushrooms can be sold for credits.
• Credits can buy upgrades.
• Upgrades increase mushrooms per click and add passive mushrooms per second.
• Show current mushrooms, credits, mushrooms per click, and mushrooms per second.
• Add 4 upgrades with rising costs.
• Make the UI readable and slightly cozy, with a lunar farm theme.
• The game should be playable immediately in the browser.
• Keep the first version small and focused.

I used a browser-based AI game maker I've been testing for this pass because I wanted the quickest route from prompt to playable link, without setting up a repo or spending the first hour wiring basic UI. The useful part was not that the AI "made a game." The useful part was that I could immediately feel where the loop was weak.

What the first version usually gets wrong

Common problems:

• The first upgrade is too expensive
• Passive income starts too late
• The player has no reason to choose between upgrades
• The numbers grow, but nothing changes emotionally
• The theme is only visual, not mechanical
• The game has no short-term goal after the first minute

A better second prompt

Revise the game balance so the first upgrade can be purchased within about 25 seconds of normal clicking. Make the early game more satisfying:

• Add a small floating +1 text when mushrooms are harvested.
• Add a visual change when passive income is unlocked.
• Rename the upgrades so they fit the moon farm theme.
• Add one upgrade that improves sell value, not just production.
• Keep the game simple and avoid adding new systems.

The main lesson

The prompt is not the product. The playable loop is the product. AI can shorten the distance between idea and browser prototype, which is genuinely useful. But it does not remove the need for taste, pacing, and restraint.

For me, the best workflow is:

• Write the game as one sentence.
• Turn that sentence into a small prompt.
• Generate the smallest playable version.
• Test the first 90 seconds.
• Fix pacing before adding features.
• Share the playable link only when the loop is clear.

That is enough to learn something. And for browser games, learning fast is half the battle.