AI/LLM Policy Updates

Hey all! I and the moderation team want to thank you for your input regarding AI showcases.
I’m grateful that we have a community that can discuss this hot button issue respectfully.

After reviewing the input, we have decided to introduce some changes and clarifications to our AI policy.

First of all what is not changing. Ziggit still is “by humans and for humans”. This means that the
existing prohibition on using AI to create posts, topics, and responses on the forum, apart from
rote translation (which may still be better accomplished using specialized tools rather than AI) is
still in force.

The changes to the policy can be seen in two parts:

Allowing Showcases Projects that make use of AI

There are several reasons for making this decision. First of all, it has never been the desire for
our moderation team to have to deeply review projects, nor police them.

However with the original AI policy we were frequently asked to review projects and make judgment calls on what was the appropriate level of AI. The second reason is that many genuinely good zig projects have started to incorporate the use of AI in their processes (e.g. Ghostty).

That being said, we recognize that there are many who don’t want to see these projects. To that
purpose we are requiring that showcase projects that use AI disclose this, and use the llm tag. If your project makes use of dependencies that to your knowledge use AI, you should use the llm-deps tag.

If regulars notice that an llm topic is missing the tag, it can be added.

Explicitly disallowing slop/vibe-coded projects

As a part of the above however, we are going to disallow projects that we the moderators determine to be slop, or vibe-coded. If there is a project that you think fits that category, please flag it under the LLM policy and the moderators will evaluate it.

Muting the llm tag

If you aren’t interested in seeing any llm related content (showcases and discussions) you can visit this link and click the bell to mute the llm tag: Topics tagged llm

Outlook

We know that not everyone will like the changes, but we hope that they provide a good compromise. Lets give it some time to experience these changes and see the different groups of topics that have the llm tag, we then can reconsider whether we need additional tweaks.

54 Likes

Thanks mod team, I think this is a pretty good solution!

4 Likes

Fixed the link, Thanks!

1 Like

Explicitly disallowing slop/vibe-coded projects is a welcome change. Will the https://ziggit.dev/faq#AI section be updated to reflect this more clearly?

5 Likes

An explicit definition of slop/vibe-coded projects is also a welcome change

I like the distinction between the llm and llm-deps tags. Sometimes it is hard to find a needed dependency that is llm-free, but llms will never touch my codebase.

8 Likes

While an explicit definition could help, actually describing the offensive behavior is difficult without being either too permissive or restrictive. Even without a definition, identification of LLM slop will probably come down to ‘I know it when I see it’

2 Likes

I doubt there is ever going to be a comprehensive and objective definition for such a thing, but is there really any great confusion on what this is?

2 Likes

Good policy.

The peanut gallery likes to paint Zig as an anti-AI cult but the reality is that by keeping Zig AI-free we just give more freedom to those who use the language.

If somebody wants to create a Zig project that uses LLMs, they can do so. And if somebody wants to create a fully LLM-free project, they can also do it.

If Zig used LLMs then every Zig project would have to be tagged llm-deps :^)

(just one thing to think about, this is not the full reasoning for wheter to allow or ban ai PRs in Zig)

21 Likes

Thanks team! The air is fresher already :slight_smile:

I understand that the policy clarification is only for the forum. But Zig itself also needs some protection in this trillion dollar AI propaganda madness. And although my first suggestion (bottom) was half in jest, the more I think about it, the more it seems like the strongest way for Zig to stay pure to:

Is there any discussion ongoing along these lines?

I mean, you are all probably writing the last human crafted compiler/language for an industry that is being bulldozed by BigTech/AI. Full respect for sticking to your values! It feels very special to me, and worthy of some kind of protected status :slight_smile:

Especially because it feels like we are racing towards a point where anyone will be able to take v1.0 and create a possibly incompatible derivative for next to no cost.

Any thoughts on how Zig will survive or handle any of this?

PS is this forum itself AI enabled? I see Discourse promote such options.

1 Like

That’s a typo on my part, It should have read

Ziggit still is “by humans and for humans”

I certainly don’t deign to put words in Andrew or the ZSF’s mouth. (For future readers, this has been corrected)

3 Likes

AI options and plugins are turned off .

5 Likes

That license change would mean that Zig would not be really open source anymore.

More in general the idea behind Zig is to give freedom to its users, not apply restrictions. This is also why Zig is MIT and not GPL.

It’s a choice that comes with pros and cons, like any other, but that’s the idea.

Se also my comment above on this general subject.

17 Likes

How far are we tracking llm-deps? :slight_smile: Zig depends on LLVM, which does permit LLM use for their changes. So every time someone is building their Zig project in release mode, they are using LLM code.

8 Likes

Fair, well that won’t be the case forever :^)

5 Likes

i see no reason to think that zig will be “the last” of anything, personally!

3 Likes

Well I do wonder how many people have the strength of character to go up against this tsunami of change and commit 10+years of pure human effort and human ingenuity into creating something as wonderful as Zig. Certainly no business would ever fund such a project proposal at this stage. So it could only be done as another labour of love.

But granted, maybe there are other programming languages that have also committed to staying human-made?

Can we also have a policy like that, if anyone heavily advocates AI here, or constantly and aggressively challenge the AI policy, or overly using sarcasm, “freedom”, or “free speech”, or quoting out of context to challenge any Zig community policies, the users will be banned, or at least their posts will be limited or tagged?

I mean, I already saw some people here doing this. And I also strongly believe that their posts and replies are partially generated by AI too. Some of my replies were instantly quoted out of context, then following replies are like a spam or something…

I hope the moderators can look at those too.

3 Likes

Sure, but when did “Users” include machines or LLM Bots?

For me Open Source was always about developers freely sharing with other developers. We are problem solvers, and like to share our solutions for the betterment of our industry, and benefit of our colleagues. Sharing with machines feels like a betrayal of the human effort.

I fight for the users! :slight_smile: