Contributing (and AI) policy

[cyrneko]

hiya!

full disclosure first, I maintain the Starlight Network No-AI list and come here in search of projects I use to gather their stance on related issues

now, continuing from that

---

Many projects have a file that describe various commonly asked questions about contributing, i.e around code-style, how commits should be formatted, getting a dev environment running...that kinda stuff.

Often it also outlines policies, like e.g a requirement to sign of on the DCO¹ or alike.

This would be a great place to put a policy regarding the use of AI² to contribute to dinit.

There is some text on justsayno.ai that goes into reasons why one may reject it and I encourage looking over it quickly. On the above-linked list I maintain you can find a plethora of projects from all 'round the software world that have already pledged that they won't use it and don't want to have it used in contributions.

This issue is, to me and many others, very important and I hope you share these concerns as well. Thank you for all your work on dinit!³

Footnotes

1. Developer Certificate of Origin ↩

2. When I say "AI", I mean LLMs and diffusion models and anything else that falls under "Generative AI". For the most part, at least. I don't count "classic" deterministic non-ML systems like Intellisense under this definition. ↩

3. I'd spare a bit to show my support but I don't got much on me right now. :( ↩

[cyrneko]

for the record i put an archive.org link for https://justsayno.ai because it seems to be down rn

[tiagozip]

why are people trying to force maintainers to adopt their ai policy and upvote brigading on it ??😭😭

[cyrneko]

I did not ask for any specific policy document to be adopted, and I made clear why I am here and where I come from

asking doesn't cost anything, and I know that many of my followers do as well so I figured I'd share it.

A few things though:

what I've written is explicitly open as to what policy actually gets adopted. If the maintainer decides after this they won't add one or adopt a different policy, so be it.

you could've at least done courtesy and provided a link to the post, or outlined how someone with twice the followers sent you here.

[tiagozip]

fair enough.

to be clear, my issue isn't with your thread itself. it's with the 'upvote this!!' post, because even if that wasn't the intent, the maintainer probably sees what looks like actual users being interested since you didn't link it, which is a form of pressure whether you meant it that way or not.

[wolfskaempf]

Thank you for bringing this up :) I’ve been working on ways to replace systemd with more focused, smaller components lately to see if I can reduce my exposure to projects that allow LLM generated text that looks like code.

dinit is one of these more focused components. It would be sad to see it go down the same route, especially as it would put projects like Chimera Linux which have stated to not accept contributions with LLM involvement into a tough position.

But the following quote from the README makes me hopeful, because the technical side of the arguments against LLMs are already part of the project’s goals and the ethical and environmental ones are easy to get behind :)

Development goals include clean design, robustness, portability, usability, and avoiding feature bloat […]

thanks everyone who makes software by humans for humans 🫶

edit: i accidentally a word, fixed