:: Re: [DNG] Debian Votes to Allow AI
Top Page
Delete this message
Reply to this message
Author: Simon Walter
Date:  
To: dng
Subject: Re: [DNG] Debian Votes to Allow AI
On 2026-08-30 14:35, David L. Craig via Dng wrote:
> On 26Aug30:2135+1000, Ralph Ronnquist via Dng wrote:
>
>> On Sun, Aug 30, 2026 at 12:51:13PM +0200, Johan Helsingius via Dng wrote:
>
>>> I wonder what this means for devuan - I assume the AI-generated code
>>> will propagate from debian into devuan. Not sure how I feel about it.
>>
>> Yes, I agree. Since Devuan only forks a few packages and serves all
>> the rest directly from Debian pools, the immediate future does look
>> like that.
>>
>> Even the forked packages will potentially "suffer" in that they
>> generally are maintained so as to offer all the functionality of the
>> upstream Debian source, other than systemd related parts. Doing
>> something else is a significant change to that work
>>
>> Thus, at this stage at least one must put Devuan on the same shelf as
>> Debian in regards to containing AI slop or not.
>
> There's another possibility. Devuan (Debian, too, if they so choose)
> can fight the slop using AI. Tell it to find the slop and then you
> as the HITL verify its findings to arrive at the truth. This can be
> handled by truly Free AI; e.g., OLMo 3, that you can fix in the
> language it was broken in, running on affordable hardware you control
> and can truly sandbox if you think that is warranted.
>
> Just be aware of the possibility if nothing else...


True true.

I want to read what they voted on. Because voting to allow AI, sounds
like voting to allow slop/vibe codez.

It is possible to use "AI" to write very concise bug free code, but that
is not vibe coding. If I were to use "AI" to write code, I would take
responsibility for it. I would not say that "AI wrote this". That would
not only be false, it would simply show that I don't know what I am
doing and cannot be trusted as a human. Sure, I make mistakes, but, I
own those and do my best to correct them. To give credit to "AI" simply
means that you cannot be trusted. If I owned a fake car and talked about
how great it was, would you trust me? It's literally fake intelligence,
which is magnitude worse than a fake car. "taking out our eyes to look
at them" comes to mind.

It seems that "AI" working on the kernel is going in the wrong direction
- allowing "AI" to make radical decisions. It's is not what I would have
expected from the Finish man.

IMO, it would be wise to wait till LLM technologies really and truly
matures, before voting on it. This tech will change so much over the
next few years that what is voted on now, may not make any sense in the
near future. I would have thought that debian's conservative roots would
have been to make "let's wait" decision.

If debian is no longer the stable linux distro, maybe more of devuan
needs to be independent, but in that very vein, to that, I'd say, let's
wait. ;)