A Game Development Attempt by a Finance Professional: Me and Codex Created Ward 09

From Finance Work to a Game Project

Hello, I'm Leon Wu, TheQuestion, the Question Pie Teacher. My full-time job is in finance, mainly responsible for internal controls, procurement, and AI-enabled finance transformation. Outside of work, I also try to turn some ideas into actual usable works as an individual developer.

Ward 09 is one of them. It's a first-person survival shooting demo: the player wakes up in a medical ward, restores communication, finds an access card, and proceeds through different areas along the mission line to escape. It's far from my daily finance work, yet it gave me a chance to test a question: how far can an individual push an idea in an unfamiliar field with the help of AI?

First, let me show a gameplay demo. Compared to describing 'making a game' with words, seeing it run is more intuitive.

It went through several changes

This project went through a web prototype, a Godot version, and then to an Unreal Engine Demo. In the early stages, we first made the movement, shooting, enemies, and missions into a playable flow, and then gradually adjusted the scenes, actions, sounds, and interface. Each stage left a playable version, allowing me to look back and compare: what exactly did this change improve?

The currently organized V1.7 is a Windows offline bilingual version, including the ward prologue and five connected campaign areas, with six types of enemies, four weapons, melee knives, and grenades, as well as supplies, a mini-map, mission hints, and short dialogues between the protagonist and the radio assistant.

The Chinese-English switch covers menus, missions, subtitles, logs, and story voiceovers. For me, the key was to let others download, launch, and understand how to play, which transformed it from 'a project on my computer' into a shareable work.

How did I and Codex divide the work?

I was responsible for proposing directions, selecting references, running playtests, and clearly stating what I didn't like. Codex helped implement systems, write Blender scripts, integrate materials, troubleshoot, adjust the bilingual interface, and prepare and verify distributable programs.

This collaboration involved a lot of specific modifications: whether the knife swing action was reasonable, whether the crosshair was blocked when aiming, whether the grenade was still rolling unnaturally after landing, whether the mission hints were clear, whether the dialogues were too long, and whether the music could change with the game state.

These issues often only became apparent after actual testing. Proposing requirements, seeing results, pointing out problems, and making modifications repeatedly happened throughout the development process. AI helped me complete many implementation tasks, and I also needed to continuously judge: does the current result truly match the initial experience I wanted?

After it runs, there's still a lot of work to do

In the development log, some modifications are unrelated to combat but directly affect whether a person can start playing smoothly: window and full-screen switching, mouse sensitivity, pause menu, language selection, installation and uninstallation, and whether it can start after leaving the development environment.

Verification also needs to be clearly defined. The existence of resources does not mean they play normally; a mission being scriptable does not mean the player knows where to go next; automatic checks passing cannot replace human listening, feel, and difficulty judgment. These boundaries became very specific in the project.

When I did finance work, I cared about whether the basis, process, and results were consistent. This habit is also useful in game development: keeping versions, recording changes, running checks, and then looking at the actual effect. After seeing a nice screenshot, I still need to confirm the parts beyond the screenshot.

This is a game assisted by AI development

Here, AI mainly participated in the development process. The released game can run offline, without requiring players to connect to Codex, nor does it rely on real-time calls to large models. The enemy behavior and mission flow in the game cannot be described as driven by large models in real time just because AI was used during development.

Similarly, this project used game engines, production tools, and materials, which also included third-party resources. Codex helped a lot with implementation and integration, but this does not mean all content was originally generated by AI. For information on materials and credits, please refer to the project's accompanying notes.

It's still a Demo

The current version can save settings preferences, but does not have campaign progress saving; the prepared version is for Windows, and a macOS playable version is not yet complete. There is also room for improvement in art, combat feel, balance, and broader hardware compatibility.

I hope you can see it as a personal development practice that you can experience firsthand. It has already crossed the distance from an idea to a playable work. How far it can go next still requires more time, feedback, and judgment.

What does this attempt mean to me?

AI collaboration allowed me to try things that were hard to push forward independently before, and it also made me value the ability to clearly explain problems. The more I progressed, the more I needed specific goals, observable feedback, and the patience to adjust repeatedly.

From professional shooting athletes, to international accounting studies and financial work, to today's AI applications and personal development, my experience has gone through several directions. Ward 09 is one interesting intersection: it allowed me to turn my interests into works and then let others test the works.

If you have watched the demo or actually played it, feel free to leave a comment below. What do you care more about: atmosphere, control feel, mission guidance, or do you want to know the specific process from web prototype to Unreal Engine? I will use these feedback as references for future sharing and improvements.

Project page: https://github.com/flyfox666/ward09

This article was assisted by OpenAI Codex in organizing based on my project records and development communications.