Why AI-Driven Game Decompilation Is Dividing Its Community

Summary

In this opinion essay, decompiler Josh argues that the recent use of large language models and agents in game reverse engineering is changing the culture of the decompilation community. He distinguishes decompilation, which reconstructs source code to match an original binary, from recompilation, which translates machine code for another platform, and from the broader goal of simply porting a game. He says decompilation projects preserve programming decisions and create foundations for fixes, mods, and ports, but are also valuable because participants enjoy the difficult process of understanding old code. The author describes how LLMs can accelerate assembly-to-C/C++ work, yet argues that they remain limited by weak code quality, incomplete context, nondeterministic guesses, unknown training data, and difficulty following project conventions. He particularly criticizes people who use agents to take over unfinished public projects, release low-quality ports, and claim credit without contributing their work back. In his account, the decomp.dev progress-tracking site was shut down because users turned it into a race and used community projects for clout, not because AI had made decompilation obsolete. He acknowledges that experienced developers can use AI selectively, including for tedious maintenance or as another perspective on hard functions, while retaining human control. The essay concludes that newcomers are welcome if they learn the craft and follow contribution guidelines, but that faster output does not replace care, collaboration, or the intrinsic value of the work.