Banjo-Kazooie: Recompiled developer says claims that generative AI accelerates PC ports of classic games are either “false” or miss the point

A software developer credited with an unofficial PC version of the N64 platformer Banjo-Kazooie has posted a long, emphatic argument about generative AI tools being used to produce rushed PC ports of classic games. DarioSamo, a member of the modder group that released Banjo: Recompiled earlier this year, says the idea that generative AI enables faster porting is either outright false or misguided. His central point is that decompiling a game, turning original machine code into human-readable code, is supposed to be an exercise in understanding and preserving how the game was made, not just a way to make it run on modern platforms.

Background: Recompilation Is a Port, Not Just a Launch

DarioSamo frames the technical work in layers. A recompilation is a type of port that uses static recompilation to automatically convert a game's code into code that behaves the same on another platform. Those recompilations sometimes rely on full or incomplete decompilation projects as sources of information, so the port can behave properly on the target platform. The changes can include enhancements for high refresh rates, widescreen output, QoL improvements and more.

The preservation argument matters because classic N64 games are not merely binaries to be relaunched. If the goal is a maintainable port, the code needs to be understood, documented and kept in a form that other developers can work with.

Development: A Video Game Esoterica Banjo Tooie Video Triggered the Complaint

A recent video from the YouTube channel Video Game Esoterica covered a Banjo Tooie recompilation project that makes use of generative AI. According to DarioSamo, via VGC, the new generative AI recomp is based on tooling devised for Banjo: Recompiled, but is not the work of the Recompiled team. He criticized Video Game Esoterica for not making that distinction clear, calling it irresponsible. In his view, the video can encourage generative AI users to keep pumping out these projects just so they get their 10 minutes of fame in the channel's videos.

Latest Progress: The Bluesky Thread and the Speed Claim

People have recently been using generative AI to automate parts of this work, but in a follow-up thread on Bluesky posted last night, DarioSamo detailed the impact of people using generative AI to knock out PC versions of older games. He said any claims that AI speeds up this tedious process are false when recompilation is involved. Developers were already quite capable of getting a recomp running in less than a day back when these tools were launched two years ago, he said.

The pattern he described has lasted for several months: a new type of person enters the scene thinking that because AI gets them over the hurdle of getting the game running, the job is basically done. These people may think they need nothing but a credit card and a subscription to an AI service. DarioSamo's answer is that getting the game running is basically the first step of the journey, not the end.

As far as recomps go, he argued, AI has not provided any sort of speedup. Making a barebones recompilation was always easy and required very little time. The difference comes in the polish and in making a good and sustainable project for the wider public, which is the part that actually takes far more time.

Donkey Kong 64: Knowledge Beat Automation

DarioSamo pointed to DK64 Rekongpiled, a port of Donkey Kong 64 that does not make use of generative AI, as an illustration. He said the Rekongpiled team were racing against another project that did use AI. Because the developers were deeply familiar with DK64, they were able to get the game to a very good state, further sped up by DarioSamo giving them information on how to use the framework to its full potential. Meanwhile, he said, the AI effort failed to make further progress because it was lacking that knowledge. He did not specify which generative AI Donkey Kong port he was referring to.

Decompilation and ROM Matching: A Result Can Compile Without Being Useful

DarioSamo conceded that generative AI can allow for faster decompilation, but only if the goal is simply ending up with code that compiles into something matching the original game's ROM file. The entire point of decomps, he said, has always been documenting the code and understanding how it works. That is why decompilation and porting efforts are usually done by different groups of people: one group works on understanding and documenting the game's code, while a different group can take that documented codebase to aid in their porting efforts.

Large language models, he continued, will often bruteforce this process and end up with a result that compiles to the correct ROM but is not readable or understandable by a human, or even by the LLM itself. That makes it harder to create both a port based on the decompiled code and patches for the port once released.

Community Impact: Collaboration Shortcuts and Demotivation

Using these tools also means shortcutting the process of working on the code in tandem with other port makers and players, DarioSamo wrote. That means new people entering the scene are deeply unaware of a lot of small but critical decisions early in development, which can lead to the tools making an unsalvageable project that needs to be restarted from scratch.

He also said the rush of generative AI ports can demotivate a developer who was in the middle of creating a project that checks all the boxes of the quality standards expected by the public. For him, it is important to show support to these developers, and allowing promotion of AI-generated projects in community spaces would discourage them further. Assuring them that there is an audience that cares about the details and things being done properly is very important.

These old and beloved games have been available in numerous ways for decades, DarioSamo added, so there is no rush to be the first to provide a solution using these tools, and it is unlikely that such a project will be the only solution in the future.

The broader concern behind the argument is preservation. If a project only produces a matching ROM, it may miss the opportunity to grasp and preserve how the game was made. In a creative sector already plagued by cyclical brain drains and the loss of crucial documentation, that is not a minor flaw. The criticism frames the behavior as deliberate, anti-social obliviousness for the sake of a notional fast return, and connects it to the generative AI philosophy at large.

Wider Modding Backlash and a Reporting Correction

The discussion around generative AI use for recompilation echoes similar anger in modding communities. Recently, the creator of a Total War: Warhammer 3 mod deliberately corrupted their creation to stop people using it for AI-related purposes. Back in August, Mark interviewed NexusMod users about what they consider an acceptable level of generative AI usage and how platforms should enforce transparency.

The original article was edited to correct a misreading of DarioSamo's post about being able to get a recomp running in less than a day back when these tools were launched two years ago.

Current Stage: Standards, Credit and What Counts as a Real Port

The dispute now sits at the community-norms stage. It is not only about whether a game can be made to run quickly, but whether a project deserves promotion when it uses another team's tooling, leaves code unreadable, stalls after the initial runnable state, or skips the documentation needed for patches and future work. DarioSamo's position is that AI has not provided a meaningful speedup for sustainable recompilations. The outcome readers should watch is whether modding spaces continue to reward projects that only need to run, or support the slower work of preservation, readable code and long-term port maintenance.