The website decomp.dev said Thursday that it is shutting down because “the community aspect of decompilation has been broken” by AI, and that projects that once required intense collaboration between people with different skills can now be done trivially with AI. The administrator of the website said that it has become impossible to keep track of which projects are primarily human-led and which ones are done by AI.
“Historically, matching decompilation has been a brutally difficult, time consuming task, on the scale of years for any moderately-sized game with a number of active contributors,” the pseudonymous admin of decomp.dev, who goes by Encounter, wrote on the site.
But recent frontier AI models like GPT-6-Astra have made it trivially easy to decompile a game: “They are significantly better at following complex directions, cross-referencing resources for symbols or debug information, and writing natural code with enough guidance,” Encounter wrote. “It is now clear that, with the right guidance, human input, and code reviews, AI can expedite the decompilation process in a significant way. What is also now clear is that the community aspect of decompilation has been broken.”
AI-powered video game decomps and recomps (recompiling code after modifying it) are everywhere. People have, for example, used AI to put Mario into Elden Ring, port games to platforms they were never designed to work on, or to combine various games in any permutations you could possibly imagine. In recent days, tools specifically designed to decomp games with AI have popped up, such as “decomp.wtf,” which promises to “point the AI agent you already use at a game you own. decomp.wtf turns it back into C, one function at a time, and checks every function byte for byte against the original.”
Decomp.dev was a collaborative website that hosted file libraries and tools for reverse engineering, re-coding, and decompiling existing video games, meaning taking an existing game and recreating its source code. Decompiling a video game can be done for a variety of reasons and is popular in a few different niches of video game fandom. Speed runners decompile video games to find potential glitches or advantages they can use to beat a game faster; hobbyists decompile video games to port them to new platforms or to emulate them (for example, making Super Mario 64 run natively on PC); modders decompile games so they can more easily modify them; archivists decompile games to recreate code that may be lost; and some people decompile games just for the challenge.
Encounter wrote that the use of AI in the community has sped things up to the point where it has become overwhelming to track human-led projects, and people using AI have taken existing, human-led AI projects and have created “hostile forks” of them, where open-source code written by humans is taken and “finished” by AI: “These kinds of hostile forks put little to no effort into meaningfully contributing to existing projects, tearing apart our community fabric.” Encounter added, “anyone with an AI subscription can manage to create a port or static recompilation that functions at a basic level […] While community members put in the effort to decompile a game, opportunists have frequently raced to release a half-baked, barely-functional port from it.”
Lemmy a year ago: “let’s go open source clones, software should be free, fuck IP, fuck Nintendo and adobe”
Lemmy now: “no, not like that”
Do you really think this or do you need someone to explain to you why this isn’t great?
Sure, explain why someone reverse engineering adobe software or older games and open sourcing it isn’t great
Because its worse code quality, also because you’re fucking up hobbiest programming.
What do you think happens when everyone runs outta money to train the corporate frontier models, and open-weight models stop being given out as loss leaders?
Great way to lose a communal skillset to time, at least when we did this with space travel we had the excuse of not doing it in the same capacity for like 50 years because of the exorbitant expense.
I mean do you know what maintaining these things by hand will be like when we inevitably have to do so? The amount of drudging through spaghetti that’ll have to happen?
What’s your solution to that? Asking Claude to “fix the problem without making new ones, pretty please with a strawberry on top?”
My comp sci teachers/profs always said FORTRAN programmers made dough because the old guys who still knew it where dying off, despite it still being used for older financial systems.
Never expected that to happen to the then cutting-edge knowledge I’ve gained both from school and self study by the time I hit my early 30s.
But at least I know I’ll probably have easy money coming to me in 3-5 years when the braindrain really hits :)
Can’t compare code quality for code that was never written. If someone reverse engineers a game that no one was working on with AI that code is better then nothing.
Doesn’t have to, I’m fully for having platforms for only humans and if someone wants to make a platform for reverse engineering without AI that’s fine by me. Nothings stopping people from playing chess even though computers are objectively better at it now.
The models will still exist. If anthropic goes bankrupt tomorrow the weights for opus 5 will still exist and be sold off to someone, it’s not just going to disappear. Same with open weight models, plenty of people have them downloaded and can torrent them worst case scenario if the companies stop distributing them.
In general I think you’re underestimating the capabilities of modern AI and overestimating the capabilities of your average developer. At this point properly made AI code is better and more maintainable then 90% of developers code.
Yeah it’ll make it so you lose skills, but any technology that helps you will. Cars make people unable to walk as far, gps makes people dependent on their phone for directions, hell even shoes stop callouses and make it so you can’t walk without them.
Even in programming we offload large parts of functionality to libraries we don’t know the inner workings of.
Stack overflow probably atrophied programmers ability to think through a problem and read manuals, but overall it produces better code because, as any SO thread will show you, people come up with stupid solutions to problems and having a source to say what is the correct solution increases code quality as opposed to a programmer thinking through it to a naive solution.
We know AI produces unmaintainable spaghetti, having LLM-generated versions reduces incentive for human programmers to make their own version.
The community mentioned in the article did that, found it was untennable, and decided to shut down over it…
That’s nice, but what happens in a coupla years when languages and libraries continue to be updated and your model isn’t trained for them? How long can you bodge together it to work with until you have to train something new? Who will handle the cost of that?
Does that still count when you talk about people who have been developing for 5 years? For 10?
Both of those are common skills that can be built up over a matter of weeks to months at most, as opposed to niche professional skills which take years to build.
The skills needed to properly keep an LLM in line will atrophy as you use them less and less.
Yeah that’s why I’ve always used as little third party code in my personal projects as possible, I’ve never been a fan of over using other peoples code.
First off, programmers have had forums to ask questions since the Usenet days. Granted, most of them never got near how big stack overflow was even 10 years ago.
These forums are double edged sword for education, yes you can copy and paste, but the best answers were posted by experienced devs clearly and carefully explaining what and why they did.
Because unlike an LLM, they understand WHY they’re doing what they’re doing.
When used properly I would argue stack overflow, like any other good documentation, can make you a better programmer, not worse.
AI decompilation is not free software. You’re relying on proprietary models, for one thing. And you’re starting from a proprietary base program. Sure, you might get something usable at the end, but “free software” means you don’t get sued for it either.
The development of the software requires propietary models, not the software itself. If I use windows with a propietary IDE to develop software and make it open source that software is still free, even if the build and development system requires propietary software.
Clean room rebuilds are legal under copyright law. If you can prove that you didn’t see the source and only made something based off observing it’s functionality that is allowed.
That’s part of the point, though. Software developed on Windows in a proprietary IDE could’ve been theoretically developed on Linux with neovim, but software developed using proprietary AI couldn’t have been developed in an open source way; there’s no meaningful way to replicate its output for confirmation. You don’t know how it was made, and you can’t produce its output for yourself. You can’t continue development in kind.
Theoretically. We’re still in a world where the corporations have control over the courts and could very easily spend enough money to get them to rule in their favor.
Besides, I think it could be very easily argued that an AI decomp isn’t a “clean room,” since you’re directly modifying the built binary.
EDIT: also worth noting, Windows and Visual Studio’s developers were paid for their work (the developers of AI’s training data were not), and Windows and Visual Studio do not have a major negative environmental impact. Not to mention, learning decompilation is a valuable skill that translates to other disciplines, and abandoning it entirely means that we’re losing that skillset as a society.