Lets build a DOOM-Inspired Game using Amazon Q

Introduction

AI agents have evolved from simple chatbots to game development partners—and I just built a DOOM-inspired 3D game with one. Here’s how Amazon Q Developer CLI surprised me (and where it struggled).

One afternoon, while scrolling online, I stumbled on an AWS challenge: Build a game using Amazon Q Developer CLI with Pygame. Intrigued, I decided to push the limits. Why settle for simple? Let’s build a DOOM-inspired shooter with complex mechanics and see if the AI could keep up.

Spoiler: The results surprised me. Here’s how it went down.

Here's how it went down.

Setting Up Our Environment

This was easy, all we had to do was install the Amazon Q Developer CLI. We would then use the capabilities of this tool to work out what else to install and how to install it, including any of the other software we needed on our quest.The easiest way to get started, I found was THIS LINK, this excellent blog by Haowen Huang is a one stop shop on how to set everything up. Q Developer CLI is a coding assistant that brings generative AI capabilities directly to the command line. While not specifically designed for game development, but remarkably effective when paired with Pygame.

I knew I had everything running by executing

Choosing the Engine: Why Pygame?

For this I decided to go with Pygame which is a a free and open-source cross-platform library for the development of multimedia applications like video games using Python. I had previously been shown what Pygame could do and thought it would be a great fit. I was familiar with Python before and installation is also simple.

The Build Process: "Vibe Coding" with AI

Time to build. I opened my terminal, fired up Amazon Q, and embraced what I call "vibe coding"—iterating with AI in real-time.

Key Prompting Strategy: Talk Like It’s a Baby

The secret? Command the AI like you’re teaching a child. Break tasks into tiny steps:

"Create a player sprite that moves with WASD."
"Add collision detection for walls."
"Implement a shooting mechanic with left-click."

This prompt chaining prevented the AI from overwhelming itself (and me).

This is where Q Developer CLI started doing all the heavy lifting for me. It created a code repo, created a README.md and started to fill out the python file shadowlabyrinth.py with code. Once complete all assertions and mechanics were outputted to the console.

The First Output: A "Crappy" Prototype. Functional but rough. The movement was janky, combat felt weightless, and the visuals were placeholder rectangles. Exactly as expected.

So after initial prompts the basic game mechanic was working. It still needed some "polishing" before it can be "playable". My next prompts were as follows:

Here’s where the real work began. My "vibe coding" principles:

Focus on game feel first (e.g., tweak player speed, gun recoil).
Build the minimum viable prototype, then expand.
Use AI to debug (e.g., "Fix the delay between shots").
Test constantly—every change impacted performance.
Iterate based on feedback (my own and the AI’s suggestions).

The Final Touch: Assets

No more rectangles! I scoured the web for royalty-free assets:

Textures: Brick walls, eerie floor tiles.
Sprites: A pixel-art gun, zombie enemies.
Sound Effects: Gunshots, groans (because what’s DOOM without groans?).

After hours of tweaking, the game transformed:

Before: A clunky grid of shapes.
After: A moody, fast-paced shooter with satisfying combat.

Final Game Showcase

Here, I uploaded the final Gameplay. You can check it out here:

For those interested in examining the technical details of the Q CLI-generated code, you can access the complete codebase in my GitHub repository.

Lessons Learned

Talk to AI like it’s a baby – Clear, step-by-step prompts ("First, create a player sprite. Next, add WASD movement") worked far better than complex, open-ended requests.
Prompt chaining is key – Breaking tasks into small, sequential steps kept the AI focused and prevented messy, overwhelming outputs.
Expect iteration (lots of it) – The first working version was rough. Playtesting and refining—movement, combat, visuals—made it actually fun.
Assets make all the difference – Royalty-free textures, sounds, and models (easily found online) transformed a basic prototype into something immersive.
AI won’t replace game design – It excels at coding help but can’t judge "game feel." Human intuition was essential for balancing mechanics and fun.
The future is AI-assisted, not AI-made – This experiment proved AI can speed up development, but creativity and polish still rely on human input.

Final Thoughts

This experiment proved AI won’t replace game developers—but it will supercharge them. Tools like Amazon Q CLI democratize development, letting small teams (or even solo makers) punch above their weight.

Want to try it yourself? Download and install Amazon Q developer CLI

Site Terms, Privacy, and more.