New Rule Forbids GNOME Shell Extensions Made Using AI-Generated Code (phoronix.com)
- Reference: 0180381317
- News link: https://tech.slashdot.org/story/25/12/14/0453227/new-rule-forbids-gnome-shell-extensions-made-using-ai-generated-code
- Source link: https://www.phoronix.com/news/GNOME-Extensions-Block-AI
> Due to the growing number of GNOME Shell extensions looking to appear on extensions.gnome.org that were generated using AI, it's now prohibited. The [2]new rule in their guidelines note that AI-generated code will be explicitly rejected:
>
> "Extensions must not be AI-generated
>
> While it is not prohibited to use AI as a learning aid or a development tool (i.e. code completions), extension developers should be able to justify and explain the code they submit, within reason.
>
> Submissions with large amounts of unnecessary code, inconsistent code style, imaginary API usage, comments serving as LLM prompts, or other indications of AI-generated output will be rejected."
In a blog post, GNOME developer Javad Rahmatzadeh explains that "Some devs [3]are using AI without understanding the code ..."
[1] https://www.phoronix.com/news/GNOME-Extensions-Block-AI
[2] https://gjs.guide/extensions/review-guidelines/review-guidelines.html#extensions-must-not-be-ai-generated
[3] https://blogs.gnome.org/jrahmatzadeh/2025/12/06/ai-and-gnome-shell-extensions/
What's next? No auto-complete as well? (Score:2)
General, sweeping prohibitions are stupid. LLMs are a tool and need to be used properly like any other tool. To forbid their use completely harms those who can use them well.
Re: (Score:2)
From mid way through the summary:
> extension developers should be able to justify and explain the code they submit, within reason.
> Submissions with large amounts of unnecessary code, inconsistent code style, imaginary API usage, comments serving as LLM prompts, or other indications of AI-generated output will be rejected."
This seems imminently reasonable. Which part do you disagree with?
Re: (Score:2)
I assume that Aviation Pete simply didn't read anything but the title.
Anyway, I replied not because of this but because of your sig "A cat can't teach a dog to bark." I'm sure this means to say something, but considering that dogs don't need to be taught to bark, I do not think it means what you think it means.
Re: What's next? No auto-complete as well? (Score:2)
For coding I could accept using AI myself to point out bugs in my own code and use AI to create custom icons instead of drawing them myself.
That's what I see as reasonable use of AI. But if you generate code using AI then you won't know what's wrong. Making good specifications us extremely hard, even harder than coding.
I know that from experience writing specifications to coders in another country. The cultural difference is causing a lot of trouble.
Re: (Score:2)
>> comments serving as LLM prompts, or other indications of AI-generated output
> This seems imminently reasonable. Which part do you disagree with?
these last 2 points seem unreasonable to me. they do not address code correctness or quality at all and would qualify as "sweeping prohibitions" on ai use like gp argues. of course i don't know if these are actual guidelines or specifications or just sloppy reporting/communication.
i would add that "inconsistent style" might be not very reasonable either (ai or not) but that's just my subjective view on an often controversial and context-dependent topic. i'm not at all against style guidelines per-se but the
Re: What's next? No auto-complete as well? (Score:2)
Hey, if your attention span prevents you from readying the whole summary, it's a real shame you still had some attention left to prove this to us
Re: (Score:2)
Except this isn't prohibiting code generated from an LLM. This is prohibiting code from someone who doesn't understand coding well enough to clean it up enough so it looks like a real person developed and reviewed it.
If the code is calling non-existent APIs or has AI prompts left in the comments, it is garbage code that shouldn't be accepted. What they're actually saying is they won't accept crappy and careless code.
So why not (Score:2)
Just make them use formal methods and be done with it?
Re: (Score:2)
People are too lazy to write a simple readme, good luck making them use anything more involved.
Script kiddies are back (Score:2)
> ... using AI without understanding ...
There's a name for that: Script kiddie. But I remember it happening before the me-too behaviour got a name. When the Windows Vista side-bar appeared, thousands of people copied the source code for a clock widget, put their name in title, changed the colour scheme, then published it on Microsoft gallery. There were thousands of widgets with identical code, including the original copyright message.
Isn't that the point? (Score:2)
"Some devs are using AI without understanding the code..." - Well isn't that the point of "vibe coding"? That's what those pushing AI tell us, right?
Re: Isn't that the point? (Score:2)
Surely you can't be a developer and not understand the code.
Re: (Score:2)
Vibe coding definitely wasn't the beginning of some developers lacking understanding. Just a force enabler on their submissions to trusted community repos.
Re: (Score:2)
but it may be the beginning of non-developers masquarading as developers
Re: (Score:2)
There are loads of developers out there who currently rely on IDE features (that have been around for at least 15 years by this point) to write code that they dont understand, or simply copy and paste from other codebases or websites.
Committing code you dont explicitly know how works is well embedded in many software development circles these days.
Re: (Score:1)
Ok, but are these people then performing their job adequately? If you are working in a team that uses code reviews, what do you do when you get a question about some part you don’t understand? It’s going to be pretty hard to get the PR approved in a well functioning review process.
Re: (Score:2)
I think you vastly underestimate how many devs there are working for companies basically independently.
Ive been working as a developer for about 20 years now, and in IT for 30 - the first time I actually joined a “team” was less than a decade ago, before then I always worked as an individual developer for small companies (who usually had 2 or 3 devs, all working independently on stuff).
I would safely say that the number of devs who have no one checking their code, doing reviews, or engaging in s
It's kind of the point (Score:2)
I think it's a viable way of "coding", but if the result doesn't compile ("imaginary APIs") or doesn't fit the requested style, then it shouldn't be submitted.
I certainly let AI code things for me for research or as utilities, without attempting to fully understand the code. I ensure it does what I want and often try to get the opinion of more than one AI. It's a nice tool for getting stuff done that's not my main work. But submitting such code would require at least a code review and ensuring it works. At