Culture Open source game engine Godot will no longer accept AI-authored code contributions - At risk of drowning in AI slop code, Godot is firming up its contribution requirements.

Godot has been suffering from a slop problem. In February, the maintainers of the open source game engine, which powers games like Slay the Spire 2 and The Case of the Golden Idol, said they were deliberating how to address a rising tide of AI slop pull requests, which had become "increasingly draining and demoralizing" for the project's code reviewers.

Today, after months of discussion, the Godot Foundation and its maintainers are drawing a line in the sand. In a blog post, the Foundation announced that Godot's guidelines for contributors will soon be amended to forbid AI-authored code, pull requests submitted by AI agents, and AI-generated text in human-to-human communication.

"It is time for us to recognize that these problems aren’t going away and therefore we need to take steps to reduce the burden on maintainers while ensuring we still have a pipeline to mentor new contributors to become future maintainers," the Godot Foundation said.

The Foundation says the pileup of Godot pull requests pending review isn't all bad: It's a sign that interest in using and contribution to Godot is increasing. But the influx of contributions authored or submitted by AI is sapping the projects' maintainers of their willingness to confront the "already tedious" work of reviewing pull requests.

"If your feedback on PRs is just being absorbed by a machine and not going towards mentoring a potential future maintainer, it becomes much harder to justify spending your free time on PR review," the Foundation said.

As the problem becomes increasingly unsustainable, the Godot Foundation says it's in the process of updating its contribution policies, focusing on "adding barriers to low-effort slop" contributions, encouraging maintainers to review code, developing new contributors into future maintainers, and crucially, requiring that all contributions come from humans who are accountable for their code—and fixing it if it fails.

"AI cannot take responsibility, and we can't trust heavy users of AI to understand their code enough to fix it," the Foundation said.

The Foundation says we can expect Godot's contributing policy to soon include explicit rejections of AI-authored code, noting that contributors should only use AI assistance for "menial things" and must disclose its use. Additionally, the Foundation will reject any AI-generated text in human-to-human communications, saying it's "a basic principle of respect"—though it says machine translations "are still acceptable" if the original text was human-authored.

"Things change every day with respect to the current suite of AI tools available," the Foundation said. "We will continue taking a conservative approach in our policies towards them, but we will re-evaluate as things evolve."
-----
Article

 

[skunt]

> was this PR written by AI
> no

this kind of gatecamping shit is crazy but also very common in open source, it's just like the COC bullshit where if you made an edgy joke 10 years ago you'd get booted. if the code is good it should get merged, nothing else matters. this isn't an anti-AI move, it's trannies shoving their agenda down everyone's throats.

 

[Megaton Punch]

Just rangeban India and Canada. The problem isn't AI code, it's mentally retarded people using AI. Just like how ten years ago, the problem wasn't dependencies and libraries, it was mentally retarded people using them.

There's no such thing as a bad tool as long as it's being used by someone who's got basic competency.

 

[peachlover669]

the trannys last stand awaiting annihilation at the hands of the hordes of browns that they helped empower.

 

[Pee Cola]

Pajeets on suicide watch.

 

[Marvin]

I submitted a largely AI PR for an open source project recently.

It was a lower level library that I was wrapping and some tests were failing. I actually discovered the issue a year or so ago before I started messing with LLMs. My original solution was that my build scripts just patched away the failing tests and continued the build. The code seemed to work otherwise and I guessed it was just some unmaintained tests.

But recently I was able to use an LLM to confirm that the test was indeed broken. I used codex (have a cheap chatgpt subscription) and had it fix the tests. I looked over everything and codex' description of the issue seemed to track. So I submitted a patch upstream.

In situations like that, I don't mind AI being a part of the loop.

Small patches, little fixes or changes to tests, documentation, etc.

But I understand being a little annoyed if someone just dumps a huge AI slop PR into your lap and tells you to read it. And if you want to ask questions about how things work, you're still talking to a dumb clanker, just proxied through some moron.

That's the thing, I understand the "if it works it works" attitude. The problem is when it doesn't work anymore. You want accountability. You want to be able to talk to the author about their design and have an intelligent dialog.

I've tried to have those conversations with AI. Even just to get some internal feedback on my own designs. It's fucking dumb and I feel retarded trying it. (Claude is better though.) When I want to brainstorm, my jeet coworkers are more useful to bounce ideas off of. And that's saying a lot.

So yeah, doing free work reviewing someone's PR but all you're doing is talking to a chatbot through a really shitty interface (Github or gitlab or whatever comments).

 

[Dante Alighieri]

But I understand being a little annoyed if someone just dumps a huge AI slop PR into your lap and tells you to read it. And if you want to ask questions about how things work, you're still talking to a dumb clanker, just proxied through some moron.

That's the thing, I understand the "if it works it works" attitude. The problem is when it doesn't work anymore. You want accountability. You want to be able to talk to the author about their design and have an intelligent dialog.

Would-be contributors using AI isn't the problem here, though they act like it is. It's the low-quality part, from jeets, who were always going to contribute garbage.

I've tried to have those conversations with AI. Even just to get some internal feedback on my own designs. It's fucking dumb and I feel retarded trying it. (Claude is better though.) When I want to brainstorm, my jeet coworkers are more useful to bounce ideas off of. And that's saying a lot.

I often ask Claude for ideas on features or improvements, and it will frequently have good suggestions.

 

[AirMail]

Going to be funny if someone forks it and then the AI version slowly becomes superior.

 

[Loris Yeltsin]

AI cannot take responsibility

So a perfect fit for the Godot Foundation.

 

[22 Rimfire]

If the outcome of this is rejection of obviously AI-driven submissions and not spiralling into a witch hunt then they’re making a sensible move.

The wrong way to use AI for code contribution looks like this: ‘grok code me a feature and submit a push request’. I can only imagine how much it must suck to be on the receiving side of PRs like that, and having policy to guard against it is sensible.

The right way to use AI for code contribution is to iterate with it as a partner, understand the changes being implemented, and fully review and test the output for yourself. Then you write the commit information because you actually understand what’s in the commit. Anything short of that is unacceptable for anything other than a personal project.

 

[Marvin]

The wrong way to use AI for code contribution looks like this: ‘grok code me a feature and submit a push request’. I can only imagine how much it must suck to be on the receiving side of PRs like that, and having policy to guard against it is sensible.

At work, one of my AI brainrotted, jeet coworkers did this to me on a project for a client.

My coworker only knows Python and he wrote a proof of concept in Python of this feature we were contracted to deliver for a client. I think AWS is footing the bill for a lot of this development, BTW. They're funding lots of companies to add bullshit AI features to their code. So company A hires company B (us) to add AI nonsense to their codebase and AWS pays for it.

At my current role, everyone knows Python because it's a jeet shop so I'm the designated "Golang" guy. Also the only white guy in the office.

Anyway, I knew I was supposed to rewrite the feature in Go to integrate it with our client's codebase. But my coworker just dumped a full implementation on me and told me to get it working. I was pleasantly surprised, "oh I didn't know you knew Go, OK". So I was working on getting it compiling, tests passing, etc. The code was a little funny but nothing unworkable. The deadline was looming and I had a few questions for my coworker. "why did you do it this way? How much Go do you know?" "oh I don't know any Go, that's entirely Claude"

Oh you motherfucker. It was too late for me to stop and write it properly myself. Which I would've been willing to do! That's my job. To do the Go stuff, among other things.

But now it was too late and I was stuck delivering a huge chunk of largely unvetted code. The only thing that kept me from just actually demanding an extension and having Dikshit (he's got a dumb jeet name, not exactly this but similarly dumb) reprimanded was that Go's type checking helps a lot with the more obvious fuckups. If it was Python I absolutely would've.

The client is currently trying to digest our PR. I'm assigned to get it working so I'm on call to debug a giant turd of a PR that I barely understand. This project has been a death march.