Space Shooter

From a rough idea to a playable first build.

I used ChatGPT to explore a game concept and create sequential build prompts, then worked through them in Cursor while directing and reviewing the output.

Existing Space Shooter gameplay recording. A demonstration of the experience, not a user-testing result.

Context: My first experiment building with AI
My contribution: Concept exploration, build direction, and output review
Tools: ChatGPT and Cursor
AI involvement: ChatGPT supported ideation and build prompts; Cursor supported implementation. I chose the direction and evaluated the result.

Problem

Could a vague idea become something I could actually play?

I wanted a small enough experiment to move through the complete build process. Early conversations explored familiar mechanics, including Flappy Bird-style interaction, before narrowing toward a space rocket game.

The documented build path

01Explore mechanics
02Choose a small game
03Structure the prompts
04Build and review

Decisions

Explore, structure, build—and review.

01/03Narrow the idea

Choose a manageable direction before implementation.

I used ChatGPT to brainstorm and compare simple mechanics. The goal was a small, buildable interaction, not a broad game specification with an expanding feature list.

The recorded game shows the rocket, obstacles, score and health indicators, with pause and shooting instructions.

Evidence boundary: the recording demonstrates the game interface. It is not a user test, performance benchmark, or evidence of an autonomous in-game AI system.

02/03Structure the build

Work through implementation prompts in sequence.

Once the concept was clear, I asked ChatGPT for step-by-step implementation prompts and took them into Cursor one at a time. This gave me a way to review the developing result rather than ask for the entire game in one step.

Documented tool handoff

01ChatGPT: structure prompts
02Cursor: implement a step
03Me: review and direct

03/03Retain direction

Give each tool a role without giving away the decisions.

ChatGPT helped with thinking and structure; Cursor helped turn instructions into code. I decided what to build, evaluated the output, and guided the next step. The experiment became a starting point for later AI-assisted prototypes.

AI support, with direction and review kept human

Objective
Build a first playable experiment from an unclear idea.
AI contribution
ChatGPT helped explore and structure; Cursor supported implementation.
My judgment
I selected the game direction and reviewed output as the build developed.
Result
A working game shown in the existing recording; no quantified speed or quality gain is claimed.

Impact

A playable experiment and a workflow to build on.

The documented outcome is a working Space Shooter game and a repeatable ideate → structure → implement → review workflow.

Delivered: The gameplay recording preserves the resulting interactive experience.
What I learned: A perfect specification was not required to start. Exploring, narrowing, and structuring the work were distinct uses of AI, alongside coding support.
Not measured: No participant tests, adoption numbers, build-time comparison, performance measurements, or verified live deployment link are documented in the available record.

Start small enough to evaluate the result.

The value was reaching a concrete interaction I could inspect—not treating generated code as the end of the process.

Back to AI Lab projects