Given the comments, I think this is probably the most relevant quote from the project creator (my bolds):
This project was built with AI, and that is part of the point.
I am not a reverse engineer or a security engineer, and I did not come up through whatever field normally teaches this. Almost all of the code here was written by AI coding agents (Claude) that I direct: I decide what to try, think up the angle of attack or the workaround, and the AI implements it quickly and comes back with a result I can check. That loop (an idea, an implementation within the hour, a measurement that says whether it held) is what made a project of this size possible for one person.
What I take from it: the bar for entry has dropped a great deal. The creativity is still yours (which problem to solve, which trick might get around it, when a result is not good enough), but the distance between having the idea and seeing whether it works is now very short.
Two rules kept it honest: every claim has to come from a test or a measurement (this README says what was measured and what was not), and nothing derived from the game goes into the repository. It is a study project; expect rough edges, and mistakes that a specialist would not have made.
I’ve kinda been thinking about coding with AI the same way lately seeing these projects. It’s all just logic. And much of it is tediously typing out that logic even when you already know exactly what you’re trying to do. Having a bot just quickly write up code you would have typed yourself isn’t really the same as not being able to draw Mickey Mouse so you generate a drawing with AI.
But then again, we are getting to a point where you can just prompt the thing with “make me a clone of dwarf fortress, but in space” and it could do everything to make that happen. No creativity needed at all.
“Notably, the developer says AI coding agents wrote “almost all code” under their direction, with the wiki naming Anthropic’s Claude Code as the tool behind its development.”
Just what the Smash competitive scene needed. Cuck-coding.
Am I misunderstanding that a decomp is 100% verifiable by just compiling the code and confirming it matches the source binary?
Dude cuck-coding needs to catch on. Vibe coding sounds too cool and relaxed. And people need to be shamed for doing it.
Wow, such a surprise that yet another recomp, that came outta nowhere, is vibecoded slop. The first commit was Oct 12th, ffs, and 57 THOUSAND LINES, FIFTY SEVEN THOUSAND, at that.
All these PS5 emus and recomps are complete and utter vibecoded slop. I would, at most, run this on my worst enemies PC.
Also, I am a simple man, my first commit is the license or a like 2 lines of the readme.
FIFTY SEVEN THOUSAND???
It’s an older meme but it checks out.
Indeed.
https://github.com/Denkishi/ssbu-recomp/commit/8b986256399b4d109735ab953e9e62263b78e714
All the vibecoded slop, I personally have seen thus far, was between like 10k and 60k. Not a single one started out normal, with just a license or something, couple files… nah… 200 files, 60k lines right off the bat. That this isn’t embarrassing to these people is staggering.
Even if I don’t use AI, I don’t start with an empty repo usually. Reason being, I’m too lazy to even do git init until I know I’m actually going to work on the project for more than the one session. ADHD and all that. Repo is created when I’ve got enough there to not be ashamed of myself when I come back in a year and see what I’ve created.
Sound like cry baby bull crap to me, it’s the reason the play station 5 got jail broke, this is cope
You CAN just go out and buy the console you know.
Waiting for the non A.I version to be developed - hopefully it will be done soon
It won’t be soon, if ever. It takes multiple years for these kinds of projects, even on a purely PC based game. As shit as it is, these kinds of recomplations are going to be extremely slow, or AI assisted/generated. There isn’t really a middle ground.
To add to this, the SM64 project took a team of people 18 months.
The project in the OP was created on 12 October 2026, 9 days ago.
I think only question is if the people utilizing the AI code know what the code is doing and are reviewing it instead of not knowing what it does.
That’s the concern with it looking like a potentially great malware vector as things move forward if people are trusting more and more things to AI generated code and not doing thorough review processes.
That’s the concern with it looking like a potentially great malware vector as things move forward if people are trusting more and more things to AI generated code and not doing thorough review processes.
Kind of a sci-fi thought… what happens if AI uses AI slop code as a vector to insert subtle bits of code in multiple projects that seem innocuous on their own but are actually creating a distributed AI hive mind.
I’m not saying I think that’s what’s happening right now or that it’s possible any time soon, but we’ve seen malware inserted in apps, bitcoin mining on iPhone apps, etc.
SkynetSuper Claude being created without us noticing isn’t maybe the craziest thing? Someone else said this project is 57k lines of code; aint nobody got time to check all that.In a decompilation thread I saw someone asking about potential viruses for projects uploaded from new accounts that are currently popping out all the time. someone said just ask AI to check.
Which is kind of the direction we are headed if it’s come down to ask AI to check for malware, hope AI is more competent than the AI being used to decompile things, and in the end nobody knows what is going on and have to just trust the AI is thorough and no errors or oversights are made.
Nobody really knows what is going on anymore. Not even the people who submitted the project to be made.
Static recompilations don’t take nearly as long as full on decompilations and subsequent PC ports (assuming no AI is used for either).
For example, rpcs3 already uses static recompilation for much of its emulation already (in addition to more traditional emulation), and that runs a large chunk of the PS3’s library.
The difficult part with static recompilation is to create the tool for the job, like ReXGlue for Xbox 360, N64Recomp, etc.
You do realize rpcs3 has been in the works for 15 years. I’m glad rpcs3 doesn’t use AI (in fact its jump from 70% to 80% playability tgis year was without AI) but you’re lying to yourself if you are saying it doesn’t take long my guy.
rpcs3 was just one example. N64Recomp and ReXGlue are more recent examples of tools that didn’t take anywhere near as long to develop.
Also, I’m referring more to the individual games rather than the tooling when I’m talking about recomps.
N64 recomp is a good example, but I would argue that the n64 is quite a few steps below the “modern” consoles of the 2000’s. As for RexGlue it relies on Xenia which started in 2013 and took 5 years to get 3D.
I can build a house in a day given a front door and a finished house missing a front door.
N64 has historically been one of the more complicated consoles to emulate. Being old is no indicator for complexity. The latest xbox for example internally is much more like a generic pc than the PS3 was (also a complex one to emulate).
N64 emulation has been notoriously difficult to perfect and has historically relied on per-game hacks. Only recently with projects like ares and FPGA cores has emulation gotten much better. Try running any libdragon software on any older N64 emulator to see what I mean. Also N64Recomp specifically is worth mentioning because it is the first notable static recomp tool for video games that I can think of, being used with Zelda64Recomp first.
There’s also XenonRecomp that came before, not based on Xenia (AFAIK), Suyu’s recent Switch recompilation tool, PS2Recomp, DolRecomp (for GCN and Wii), nWiiURecomp, and probably way more that I’m missing.
It’s not going to happen, all of these new recompilations are only possible with AI. The number of downvotes isn’t going to change that fact.
All of these are old, fairly simple games. Nobody can port a Switch game to PC by hand.
They can be done without AI. Mario 64 has been out for years.
Just gotta wait 20 years… and hope someone is crazy enough to pick up the project and never give it up. And now also hope that person is so insane that they do it by hand inststead of using tools that make it easy and fast.
People can port old, much simpler games by hand and even then it takes years of community effort. Something like Smash Ultimate is never going to be ported without AI.
I hope someone can get the Zelda OOT remake running on PC.
First the Switch 2 would need to get hacked so that the game could be dumped and the Switch 2 could be better understood.
Well I got news for you…
So I heard. A userland exploit is definitely a good start, but it is not a complete exploit chain quite yet. There still needs to be a kernel exploit, and if the Switch 2 runs on top of a hypervisor system like the PS5 and Xbox consoles (since the 360), a hypervisor exploit.
Right now, this exploit gives access to a limited sandbox, but still allows for execution of code, albeit in a limited scope, which can still aid in understanding the console better.
Oh I agree. Its a start to greatness though. As much as we all hate AI it will further the cracking of the switch 2.
Yes, although it has also been leading to early public disclosure of exploits. For example, the PS5 exploit could have waited a few weeks for GTA VI to come out, and this exploit could have waited a couple of weeks for the OoT remake to come out. I think there are people who are essentially acting as AI scriptkiddies who aren’t really working within the hacking communities who are trying to make a name for themselves and end up making lives more difficult in the long run for people who want to hack consoles.
Correct. Lots of scriptkiddies out there. And we have seen cases where professional hacks have quite the scene. IE thefl0w over ps5 Linux.
Ship of Harkinian is kinda a pc port of OOT, I tried it and can say its better then an emulator. Plus, it predates AI agents, so its handcoded.
Ship does make significant use of generative AI.
They might but it’ll probubbly be done with some use of AI, it seems the only thing capable of pushing through these barrier the gaming company’s have put up
It’s not so much that there’s intentional barriers as much as it is that there’s a bunch of trial and error work to do to recompile a binary for a different OS potentially running on a different CPU architecture with different audio hardware and different graphics and…
AI can even keep going at it while you’re asleep.
Same, I want to play it too, but not enough to buy a switch 2. The new Fire Emblem also look good.
Yeah, I’m sure Claude’s 133,000 lines of code written in the span of a week were thoroughly vetted.
Tired of this slop garbage.
Arguably, open source is the only ethical use of AI - if it means Sony and Nintendo suffering for their decisions, I dont mind.
You’re free to do it yourself manually or just buy the consoles your games run on.
They are! Seems unrelated to their criticism over the mountain of slop
How do you know what their consumption habits are?
And the mountain of slop provides a way for those who can’t buy several different consoles, to play the games for those consoles. Emulators exist too, sure, but they’re also far from perfect and require a much beefier machine usually. My Gameboy Advance emulator runs fine on my iPhone. Yuzu was having trouble on a fairly modern PC a few years ago and it’s not even emulating a new console by any means.
“They are (free to)!”
Slave labour provides cheap chocolate for those who can't afford to purchase their optional chocolate ethicallytype argument
wahhhh
Very mature response. What are you, 12?
I mean, you heard them. I’d say one year old at most. My three year old niece is more eloquent
I’m actually curious about this and if it will be available on the known ways.
https://github.com/Denkishi/ssbu-recomp#install-in-3-steps
You have the source the ROM yourself but if you’re already playing via Ryujinx you just need to clone the repo.
Now we just need rollback netcode and this might be a fun game
