Planet Orbit — AI-Assisted Game Dev

Summary

A prompt-to-App-Store experiment

Planet Orbit started as a question, not a brief: could AI close the gap between a fleeting idea and a shipping product? I wanted a modern take on the Atari classic Asteroids — a ship orbiting a planet, with a twist: make it just hard enough that players feel that "just one more try" frustration.

Normally my process is sketch, wireframe, mock, then code. Here I inverted it — prompting a playable prototype straight out of Google AI Studio, then letting real players (not my own assumptions) drive every decision from there through a Firebase backend, a global leaderboard, and an actual App Store launch.

01 — The experiment

Inverting the usual design process

Skipping wireframes to get straight to the "feel" of the physics. The goal wasn't clean code yet — it was to put the device in a user's hand immediately and test the core loop.

The workflow shift

Instead of weeks between an idea and something a player could touch, the loop compressed to minutes.

Traditional process

Sketch → Wireframe → Mockup → Code
(weeks)

AI-first process

Prompt → Playable Prototype → User Feedback
(minutes)

02 — Process

Four phases, driven by what players actually did

Each phase started with a prompt and ended with a decision forced by real usage — not a spec.

Phase 1 — the 10-minute prototype

I described the physics I wanted to Google AI Studio. Within 10 minutes I had a working prototype — significantly faster than my usual workflow.

User: Create a game using HTML Canvas where a ship orbits a planet. It should feel like Asteroids but with gravity.
Gemini: Here is a working prototype using React and the Canvas API… (generated in <10s)

v0.1 — first playable build running on device.

The first playable build. Screen recording of the v0.1 prototype, before any user testing had shaped the controls.

Phase 2 — the control pivot

Listening to users to find the "fun."

I have to look at the bottom of the screen to use the slider, so I crash.
Iteration 1 — Slider (Failed)

A user suggested a "Flappy Bird"-style interaction: tap and hold to orbit, release to drift.

Okay, this is actually addicting now. I can't stop playing.
Iteration 2 — One-tap control (Success)

Early slider-control iteration, requiring the player to watch a slider at the bottom of the screen while steering the ship — Iteration 1 — a bottom-of-screen slider. Players had to look away from the ship to use it.

One-tap, tap-and-hold orbit control that replaced the slider — tap to orbit, release to drift — Iteration 2 — tap and hold to orbit, release to drift. The control that made it addictive.

Phase 3 — scaling up

The group chat effect

Once the tap controls were finalized, adoption spiked. Four friends started a group chat solely to update each other when they beat a high score. That organic competition was the signal I needed — the game wasn't just a prototype anymore; it needed a global leaderboard.

The backend problem

A global leaderboard required a real database. This was over my head — I felt like I'd bitten off more than I could chew. I used AI as a tutor, not just a code generator, to understand how to structure a NoSQL database in Firebase.

"Step 1: Create a Firestore collection named 'scores'…"
AI-guided implementation

Global leaderboard screen showing player rankings, built after friends started a group chat to compare high scores — The global leaderboard — built after a group chat made the demand for it impossible to ignore.

Phase 4 — retention strategy

After the initial hype, Daily Active Users began to dip — a 20% decline. I needed reasons for players to open the app even without time for a long session.

Daily Login Bonus — a simple streak system: log in to claim currency. This stabilized the DAU dip immediately.
The Supply Depot — a shop to spend points on cosmetic ship skins, giving players a long-term goal beyond just high scores.

Supply Depot shop screen for spending points on cosmetic ship skins — The Supply Depot — a points shop that gave players a reason to come back beyond chasing a high score.

03 — Built and shipped

What started as a prompt became a shipped product

Planet Orbit home screen, as shipped to the App Store — The finished home screen, as shipped to the App Store.

10minidea to first playable prototype

3.5kApp Store views

104total downloads

<1motime to market

04 — Reflection

Key takeaways

Rapid validation

By skipping high-fidelity mocks and jumping straight to a working prototype, I tested complex physics mechanics in minutes rather than days.

Deployment success

I successfully navigated the App Store review process, shipping a polished product to production.

Skill acquisition

I learned Firebase, React Native concepts, and backend logic by using AI as a personalized tutor throughout the process.

Play it

Planet Orbit is live — download it on the App Store ↗.