AI-assisted recomps and decomps - where do you draw the line?
You'd have to be living under a rock to have not noticed the recent boom in static recompilation and decompilation projects - it seems like every day, multiple projects are announced; from popular titles to more obscure games with niche audiences; at the rate they're appearing it feels like it's only a matter of time until every game is playable on every platform.
For example, just recently The Legend of Zelda: The Wind Waker has made its way to Windows and Mac natively - bringing with it high frame rates, widescreen support, improved rendering, HD texture pack support and more... AstroCore has brought Super Mario Galaxy to PC, gen1recomp has added FireRed and LeafGreen, the list goes on.
It's no surprise to anyone that this boom in projects has been almost fully in part to the rise of AI assisted development or in some cases, projects that are entirely vibe-coded.
Many of these projects disclose outright the fact they were built with help from large language models such as Claude Code. For example, The Wind Waker recomps readme says it was developed "with substantial AI assistance for code, testing, documentation and debugging".
So here's the question, beyond the ethical and environmental impacts; these recomps and decomps cost nothing. They’re made in part by people in their spare time and can provide easy access to games that are perhaps stuck on dead or hard to find hardware, run or perform badly on their original platforms; or that the original publisher may have delisted entirely. Yet as soon as an LLM shows up in the credits a big part of the reaction devolves to "AI = bad", whatever the resulting product may look like. Of course, some of that criticism is absolutely well founded: people worry about undocumented, unoptimized and spiraling codebases; about the craft and passion of hand reverse-engineering and about how these models were trained in the first place.
So who's really right? Is it the people who see AI-assisted ports as a shortcut that cheapens the scene, or the people who think a free, working, improved and accessible version of a game they love is worth having however it was made?
Where do you personally draw the line?
Back in August, Scarlet's community check-in on how we cover AI-assisted projects garnered more than 360 replies. Scarlet suggested that when a project we feature uses AI and makes it clear that it does; we will endeavor to inform you depending on of course on the information publically available and to the extent that you should expect from an independent website providing simple news coverage. Giving anyone who would rather avoid these projects a clear indicator, without anyone having to dig through repositories for hints themselves.
So let us know:
- Where do you draw the line on using LLMs in such recomp and decomp projects?
- Would you play a game knowing much of its port was written with AI? Or would you prefer to avoid it?
- If an AI-assisted recomp were the only way left to play a game you cherish would you take it? Or would your principles keep you waiting for an official remaster?
