One of the largest video game decompilation communities has decided to shut down its central hub, and it is pointing the finger squarely at AI. decomp.dev has begun sunsetting its website, with administrator @encounter saying the rise of code-focused AI models has broken the community side of reverse engineering, reports Engadget.
The site was built to gamify a slow, painstaking craft, letting members watch progress bars and leaderboards tick upward as a game's code was pieced back together. According to @encounter, that motivation has been replaced by speed, turning decompilation into what he called a "hollow numbers game." The main project index and the ability to add new projects are now disabled, though existing projects, individual pages and historical data will still be tracked.
What decompilation actually is
Decompilation is not emulation. An emulator recreates the environment a game originally ran on, while decompilation takes a game's compiled executable and rebuilds its original source code, typically written in a language like C++. It is a process of trial, error and deep programming knowledge, and it is far more complex than simply getting an old ROM to boot.
The payoff is control. Once a game's codebase is readable, developers and modders can improve frame rates, remap controls, fix bugs and make fundamental changes. The practice also matters for preservation, especially as the industry drifts away from physical media. A decompilation of Animal Crossing on the GameCube, for example, lets the in-game calendar run past 2030.
It is also a research tool. Full code gives curious players and speedrunners a look at hit boxes, random number generation and the design decisions original developers left in the build. As @encounter put it, decompilation communities spent years arguing that the craft exists primarily to understand the games people love.
How AI split the community
The administrator was careful to say AI itself is not the enemy. Used properly, with the right guidance, human input and code reviews, these tools can be enormously helpful. The problem, he says, is a subset of users he calls "bad actors" who have exploited the open source nature of the projects.
Much of the damage comes from forking, where someone copies an in-progress decompilation project and races to finish it with AI. @encounter describes many of these as drive-by contributions from people who put no effort into quality. There has also been a surge of AI-crafted ports built on decompiled code by people who never contributed to the original work, producing results he calls half-baked and barely functional.
