Rust and Zig now have anti-LLM contribution policies.

Rust and Zig now have anti-LLM contribution policies.

Rust和Zig现在都有反LLM贡献政策了。

Any thoughts?

有啥想法？

图片: https://i.4cdn.org/g/1786018648462739.jpg

>>109477883

Not new for zig also by the pic another OP posted a few hours ago rust apparently took a more varied stance on AI contributions

Zig那边不新鲜，从几小时前另一位楼主发的图看，Rust对AI贡献的态度似乎更多元。

Don't care, I'll stay at C

无所谓，我还是用C。

>>109477916

I’m asking if we’re going to see this more and more.

我是在问，这种事会不会越来越常见。

We know open-source is run by real programmers, and I think the tolerance for slop is at a breaking point.

我们知道开源是真正的程序员在跑，我觉得大家对垃圾代码的容忍度已经到临界点了。

>>109477883

They are valid, but five years from now, I don't think they will have much choice.

这些顾虑是合理的，但五年后，我不觉得他们还有多少选择余地。

Most rust developers are really subpar. They are terrible. I have the actual development. That's why they need rusts hand-holding in the first place. Replacing these fuckers with an artificial intelligence is a breeze.

大多数Rust开发者水平真不行，烂得很。我实际搞过开发，所以他们一开始就需要Rust那种手把手式的保护。用AI换掉这帮人简直轻松。

To me it's an admittance that you don't care about engineering which is bad for an engineering related project.

对我来说，这等于承认你不在乎工程本身，这对跟工程相关的项目来说很糟。

These are political decisions, not technical ones.

这些是政治决策，不是技术决策。

Nobody cares who writes code, they care if the code works. And most maintainers don't even write most of the code so I don't know why they would be upset to begin with, your job already majorly consist of reading code written by other people, but somehow it's okay to take their contributions but not when its mechanical. I guess improvement isn't worth it to you and your users compared to your personal pride in humanity.

没人在乎代码是谁写的，只在乎代码能不能跑。而且大多数维护者自己都不写大部分代码，所以我不懂他们一开始有啥好不爽的。你的工作本来主要就是读别人写的代码，但不知为啥，接受别人的贡献行，接受机器写的不行。我猜对你和用户来说，改进还不如你个人对人类的自豪感重要。

I'm being serious, this is effectively like putting a "we don't take things seriously here" sign on the front of their projects/offices. They would rather bicker, with machines, than allow them to contribute and ease the burden. Something you would think open source developers would care about considering they're underfunded and understaffed, yet here they are limiting their own options even further for no good reason. They'd rather have an unlimited stream of issues that they're solely responsible for solving even though they don't have to, stalling up progress for the entire ecosystem.

我是认真的，这基本上就是在项目/办公室门口挂个“我们这儿不认真做事”的牌子。他们宁愿跟机器吵，也不愿让机器来帮忙分担负担。开源开发者不是一直喊缺钱缺人吗，结果现在还自己给自己添堵，没半点理由。他们宁愿面对处理不完的issue，全自己扛着解决，哪怕本来不用这样，拖着整个生态系统的进度。

>>109477883

>>109477979

it doesn't matter

无所谓

who's stopping you from making llm code, contributing it and saying yep i wrote this? nothing

谁拦着你写LLM代码、提交上去然后说“这是我写的”了？没人。

the only thing they care about is not having to review slop the people who submit it don't even understand, you have to be fully aware of and confident in the technical decisions, if you are it doesn't matter if a machine wrote the code

他们唯一在乎的是不想审查那些提交者自己都看不懂的垃圾，你得完全清楚并自信于技术决策，如果你做到了，那代码是不是机器写的根本不重要。

>>109477883

rust really is the tranny language

Rust真是个娘炮语言

>>109478034

>it doesn't matter and you can't tell the difference

>无所谓，也分不出来

>lets enact policy that prevents this

>然后搞政策来禁止这事

This implies it's not enforceable consistently which usually ends up only burning innocent people, just like with DRM.

这意思就是没法一致执行，最后往往只会误伤无辜的人，跟DRM一样。

A human submits a patch, someone accuses them of AI, there's no way to prove or disprove it. Nobody wins.

真人提交补丁，有人指控是AI写的，谁也证明不了或反驳不了。结果就是双输。

>>109477883

>nooo you have to write code the way we want you to

不不，你必须按我们想要的方式写代码

why are communists and transfolx so authoritatian?

为什么共产主义者和跨性别者这么专制？

>thoughts?

LUDDITES BTFO

AI CHADS WON

>>109478034

this, also no one wants/can maintain slop. ai write 10x more code than necessary, and the result is some unreadable mess. maybe if you need to ship garbage in your day job it's fine, but if we're talking decade long projects with long term support, slop won't cut it.

说得对，而且没人想要也没人能维护垃圾代码。AI 写的代码是必要量的十倍，结果就是一团没法读的烂摊子。也许你日常搬砖赶着上线无所谓，但要是谈十年期、要长期支持的项目，垃圾代码根本撑不住。

it's not an anti ai decision, it's just setting the bar for quality. i read rust ai guidelines and they explicitly state you're not disallowed from pushing ai generated code.

这不是反 AI 的决定，只是设定质量标准。我读过 Rust 的 AI 指南，里面明确说了没禁止提交 AI 生成的代码。

>>109477949

what makes you say that? most rust projects ive seen are very well written

你凭什么这么说？我见过的大多数 Rust 项目都写得很好

tauri, zed, alacritty, wezterm, bevy, helix, niri, tree sitter, iced/iced gui apps, all come to mind

tauri、zed、alacritty、wezterm、bevy、helix、niri、tree sitter、iced/iced GUI 应用，都能想到

LUDDITES BTFO

>>109477883

None, thanks for asking.

没有，谢谢关心。

>>109477883

tranny lang

变性人语言

I think they are right for most of the contributions, but if someone also reads the code and has a functioning brain, AI can be useful.

我觉得他们对大部分贡献的判断是对的，但如果有人也读代码而且脑子正常，AI 还是能派上用场。

>>109477883

Rust is 'AI allowed if you look at the output' (in practice it just means blatant slop is allowed). Zig is the only serious language.

Rust 是“只要你看了输出，AI 就允许”（实际上就是明目张胆的垃圾代码也放行）。Zig 才是正经语言。

We need an anti-slop OS and an anti-slop browser, though, otherwise we're screwed.

我们需要一个反垃圾的操作系统和一个反垃圾的浏览器，不然就完蛋了。

>>109479809

Nobody who has a functioning brain gets any productivity out of gay eye. Maybe 10 years from now that will be the case and by all means, policies should be revisited then. But that simply isn't the current reality.

脑子正常的人不会从 gay eye（AI）那里获得任何生产力。也许十年后会是那样，到时候当然应该重新评估政策。但现在根本不是现实。

good

>>109478034

the fact that people can immediately identify slop generated code the same way they can immediately identify slop generated art

事实是，人们一眼就能认出 AI 生成的垃圾代码，就像一眼认出 AI 生成的垃圾图一样

>>109477979

>these are political decisions, not technical ones.

>这些是政治决定，不是技术决定。

no, they are definitely technical decisions because developers are getting completely overwhelmed with dogshit pull requests eating up their time and the code they’re getting doesn’t meet minimum quality standards and the brown proompter who one shot it didn’t even bother to read it or understand how it works.

不，这绝对是技术决定，因为开发者已经被一堆狗屎 pull request 淹没了，浪费大量时间，而且拿到的代码连最低质量标准都达不到，那个随手甩个提示词的哥们儿连看都没看它一眼，也不懂它是怎么运作的。

AI labs will simply fork Rust and keep developing it. Snailcats can have their toy for all they care. You can't stop progress.

AI 实验室干脆 fork Rust 继续开发就是了。蜗牛猫们爱玩他们的玩具就玩去吧。你拦不住进步。

>>109480295

>developers are getting completely overwhelmed with dogshit pull requests eating up their time

>开发者已经被狗屎 pull request 淹没，浪费大量时间

This is a combination of observation bias and neglecting the current state.

这是观察偏差加上忽略了现状。

We already have the problem of people submitting bad PRs. Even if the volume is greater now you're neglecting to acknowledge the increase in accepted patches while acknowledging the increase in rejected patches.

我们本来就有烂 PR 的问题。就算数量现在更多了，你光说被拒的 PR 增加了，却不承认被接受的补丁也在涨。

>>109480321

BASED

LUDDITES BTFO'D BY ALMIGHTY AI CHAD

...AGAIN!

LUDDITES ARE WEAK

AI IS STRONG

THEREFORE

KNEEL, WEAKLINGS, KNEEL BEFORE MIGHTY AI MASTER RACE

图片: https://i.4cdn.org/g/1786040176758055.png

>>109477883

Based. The slop must stop

正解。垃圾必须停掉

>>109477979

No, it is an engineering decision.

不，这是工程决定。

LLMs spit out garbage, unmaintainable, and frequently outright incorrect code, and they create PRs at a rate that is difficult to process, while drowning out actual quality contributions

LLM 吐出来的是垃圾、没法维护、经常直接是错得离谱的代码，而且它们提交 PR 的速度让人根本处理不过来，同时把真正优质贡献淹没掉

>>109481584

>sometimes the tools are ineffective, especially the older ones, so we should blanket ban them since I don't think anyone could use them effectively

>有时候这些工具不好使，尤其老的那些，所以我们应该一刀切禁用，因为我觉得没人能有效用它们

No.

>>109480014

Java (openjdk specifically) also has a kind of anti-AI policy. But the circle of people who can contribute to Java and JVM development is extremely limited and difficult to get into.

Java（具体说是 openjdk）也有类似的反 AI 政策。但能参与 Java 和 JVM 开发的人群圈子极小，门槛极高。

>>109481626

>some people survive getting shot in the foot, so we should all continue shooting ourselves

>有人脚中枪没死，所以我们全都该继续对着自己开枪

All that applies to even the newest, cutting edge LLMs. It's slop all the way down

这套说法连最新最前沿的 LLM 都适用。从里到外都是垃圾。

>>109481724

That argument isn't as convincing as you may have thought.

那个论点没你想的那么有说服力。

Most people understand the merits and utilities of a firearm even if you could decide to shoot your foot with it.

大多数人懂枪的好处和用途，哪怕你确实可以选择拿它崩自己脚。

>>109481626

I'm a different anon, but the massive influx of AI-generated pull requests means that as a project maintainer, you're forced to:

我是另一个匿名的，但海量 AI 生成的 pull request 涌入意味着，作为项目维护者，你被迫面临：

1) completely abandon reviewing these pull requests and merge everything without looking at them (rarely seen)

1）彻底放弃审查这些 pull request，啥也不看直接全合并（极少见）

2) do reviews and merges entirely using AI, without even looking at the pull requests (like OpenClaw which is unmaintaiable piece of shit after 8 month of development and already completely forgotten)

2）审查和合并全用 AI 搞定，连看都不看一眼 pull request（比如 OpenClaw，开发 8 个月后成了一坨没人能维护的屎，现在早就被忘干净了）

3) be forced to limit AI usage and start banning people, sometimes unfairly

3）被迫限制 AI 使用，开始封人，有时候封得并不公平

A lot depends on the project itself and the community. I work in the JVM tech stack, and there are many jvm projects where AI use is fully permitted or encouraged. But there's simply no massive influx of pull requests, as the projects themselves are quite mature, typically used in enterprise environments, boring, not cool. There's no hype.

很多时候取决于项目本身和社区。我在 JVM 技术栈干活，有很多 JVM 项目完全允许甚至鼓励用 AI。但根本没什么海量 pull request 涌入，因为这类项目本身相当成熟，通常用在企业环境里，无聊、不酷。没有任何炒作热度。

At the same time, projects in Rust, Typescript, and Python attract a huge number of vibecoders, and nothing can be done about it. The same situation applies to open-source game engines.

与此同时，Rust、Typescript 和 Python 的项目吸引了大量 vibe coder（凭感觉瞎写代码的人），对此毫无办法。开源游戏引擎也是同样的处境。

So, unfortunately, we can only hope that after a while, vibecoders will get bored and move on to other things, leaving open source behind forever.

所以，很不幸，我们只能指望过一阵子这些 vibe coder 玩腻了去搞别的，把开源永远抛在脑后。

>>109481736

Some firearms are good and useful, just like some tooling is good and useful

有些枪又好用又有用，就像有些工具又好用又有用一样

And then there are Sig P320 and LLMs respectively

然后分别对应的是 Sig P320 和 LLM

>>109477979

>Nobody cares who writes code, they care if the code works.

>没人在乎谁写的代码，只在乎代码能不能跑。

Programmers need to accept that these are NOT mutually exclusive for the foreseeable future as long as the automated entity is LLMeme based.

程序员得接受一个事实：只要那个自动化实体还是基于 LLMeme（LLM 梗）的，这两件事在可预见的未来就不是非此即彼的。