I wanted to see if ChatGPT could get game modding set up so my local LLMs could eventually take over. We started with script changes in World at War and Black Ops 1, then got into importing perk machines and figuring out why games would not run properly in emulators. These are my notes on how that went.
I asked ChatGPT to investigate why Local Zombies would not start in Xenia. Its experimental session changes reached TranZit and five DLC maps. The tests identified missing emulator behavior; a finished fix is still pending.
I gave ChatGPT tools to control my PS4 and test shadPS4 on Ubuntu. It reached Kino der Toten on the console. The emulator investigation reached an incomplete Local menu and a focused voice-read regression. I paused it at the September 8 checkpoint; emulator Zombies gameplay remains unverified.
ChatGPT reads saved screenshots and logs through the project tools. Moonlight and the FFmpeg stream let me watch the sessions; Chiaki carries PS4 input.
How to read these notes
I’m describing my experience using these models on my own projects. The technical details come from the saved tests and project notes, and the screenshots are from the games. Some things worked, some took a lot of retries, and some are still unfinished. I’ve kept those distinctions in the write-ups.
These are my projects using ChatGPT through Codex. I set the direction and review the results; ChatGPT writes code and runs automated tests. These pages were also written with ChatGPT.
Unofficial projects. Game names and imagery belong to their respective owners. Questions or corrections