I’m still sad that Debian is now integrating (external?) LLM in the build process which seems antithetical to the core values of Debian (as pointed out).
There is no alternative to Debian for me. I’m ready to justy throw out all my computers and become a goat herder at this point.
PS I’ll remove memory and storage from said computers to pay for land and goats before I chuck them out of my window.
I don't think there's another way. Everybody is being spammed with Gen AI outputs and the only even to it seems like a paradox way to fight it is with LLMs of some kind or go around it like what the EU is trying to do with watermarking and then detect AI and have different processing for it. I kinda think we could end up there, but we are still not there.
voidUpdate · 2026-10-27 10:54:05 UTC
The other way is not to spam someone with LLM-generated outputs, and actually review the package
monsieurbanana · 2026-10-27 10:56:42 UTC
Does debian's NEW queue actually have this issue? Not anyone can just upload to it, you have to be a debian developer or maintainer AFAIK.
tempfile · 2026-10-27 11:07:59 UTC
How can the solution to LLM slop be sending more LLM slop back?
tommica · 2026-10-27 10:49:50 UTC
> spend my valuable developer time (this is my free time)...
Lesson for us all is that if we do not own the thing/project/etc, things can suddenly change without our decision, and all time we put into it might become a waste-of-effort retroactively
lukeify · 2026-10-27 11:00:42 UTC
Hate to say it but this is one of the primary reasons I can never get interested in contributing to open source.
throw-qqqqq · 2026-10-27 11:14:55 UTC
Doesn’t this also apply to closed source software, but with the extra downsides that you can’t fork it etc.?
I contribute to open source software because it improves stuff I use and/or care about.
It seems too pessimistic, IMO, to abstain because things could potentially change to the worse, sometime in the future.
sampullman · 2026-10-27 11:22:00 UTC
Yes, I think they're pretty much saying they don't want to contribute to projects they don't own.
throw-qqqqq · 2026-10-27 11:34:52 UTC
To be fair, it is often a thankless job to contribute, especially if the project is popular (and thus get lots of low-quality PRs).
Patience and persistence helps, but it can be daunting IMO
poszlem · 2026-10-27 12:06:12 UTC
Contributing to closed source usually involves getting paid for your work.
throw-qqqqq · 2026-10-27 12:55:15 UTC
Sure, I see that point.
I replied as I did, because lukeify implied that lack of ownership is
> one of the primary reasons I can never get interested in contributing to open source
calvinmorrison · 2026-10-27 14:56:30 UTC
don't let it stop you, but don't conflate open source with open core or 'democratic governance' etc.
Open source is open source. publishing code.
it does not mean
- you need to respond to peoples PRS
- it does not mean you need a github
- it does not mean you need to provide ANYTHING to the user. No warranty.
- you do not need a mailing list
- you do not need governance
Do what you want, have fun, if you have peers or other people interested in your area, feel free to collaborate in the same way you would with pickup basketball at the park, play, have fun, and leave when you're done.
jewel · 2026-10-27 12:53:44 UTC
Debian is a special case (probably even among open source projects) because it has a well-developed system of governance:
It sounds like in this case someone is not following an agreed-upon policy. And of course, even with governance like this one, the rest of the maintainers can move in a direction you don't agree with.
monsieurbanana · 2026-10-27 10:52:18 UTC
I'm not anti-ai but I am firmly anti-make-people-read-llm-output-I-didn't-read-myself, and this breaks this.
They also have a reward system that punishes people for ignoring the llm review (by pushing their submission lower in the queue).
SOLAR_FIELDS · 2026-10-27 11:00:36 UTC
Yeah AI is the relevant stand in but ultimately this is about one group of people being lazy and trying to shift more work onto another group of people. It just so happens that they are using AI to do it
account42 · 2026-10-27 11:09:39 UTC
I don't think it "just so happens". AI enables that behavior at great scale and arguably even encourages it.
agwa · 2026-10-27 12:34:41 UTC
You can disagree with their approach, but I don't think it's fair to call the DFSG team, which is comprised of unpaid volunteers, "lazy". They have to vet every brand-new package that enters Debian, and their queue has an infamous history of being a bottleneck in the Debian project. Their time is certainly more valuable than the the time of the individuals uploading new packages.
thesuitonym · 2026-10-27 13:44:03 UTC
I understand your confusion, but here's the truth of the matter: None of this has to be done. If anyone on the DFSG team does not want to do this work the answer is simple: Just don't do it.
solyin · 2026-10-27 13:54:54 UTC
How well is that working out for the chores at your house?
monsieurbanana · 2026-10-27 11:03:05 UTC
I guess I have to temper my own statement a bit. I do see situations where I would generate and send automated LLM messages. Probably to reply to generated unsolicited messages or pull-requests.
But to upload to debian new's queue you already need to have pre-approval, it's a different situation from e.g. a public git repository that anyone can push to.
g-b-r · 2026-10-27 11:19:22 UTC
> where I would generate and send automated LLM messages. Probably to reply to generated unsolicited messages or pull-requests.
Why would you use an LLM instead of a fixed message?
monsieurbanana · 2026-10-27 11:27:23 UTC
To help classify between human and llm-generated ones? I don't know, and you're probably right than in this case a fixed message would be better.
Another situation where I would use llms to generate messages is if we've mutually agreed to do that. For example automated review messages at work, if we've agreed that it would be a part of our workflow.
A good rule of thumb might be unsolicited vs solicited llm messages.
g-b-r · 2026-10-27 12:23:39 UTC
Ok, for unattended replies indeed a fixed message is much better; you can't predict what an LLM will output, and a certain impersonal message is better than a personalized one that might be completely senseless.
If you mutually agreed of course it's fine
tetha · 2026-10-27 13:22:45 UTC
And even with a mutual agreement - default LLM output tends to be very hard and inefficient to read. We've spent quite a bit of time at work to setup a few writing skills for Claude.
For example, mine is a technical writing skill based on the old book "The Art of Scientific Writing": Front load information as much as possible - the first paragraph of a document should be a summary, the first sentence of a paragraph should summarize. Short sentences building on each other. Always show evidence.
Someone else spent quite a while optimizing a standardized PR description template, with a section for intent, expected changes, risk, rollout coordination and such.
That made the slop a lot more tolerable.
InsidiousWombat · 2026-10-27 13:21:59 UTC
But there is also way to throttle that. Have AI go thru submission guidelines, compare it to PR and have it shut up when most (not all! AI will always give a reason if you ask it to check something) of it is correct, and only answer when the PR itself doesn't comply to the project rules
pantalaimon · 2026-10-27 11:09:24 UTC
Do you share the same opinion about script generated (pre-LLM) review comments?
monsieurbanana · 2026-10-27 11:23:34 UTC
It's disingenuous to compare LLM outputs to script generated ones. They have more differences than similarities.
A script is reproducible, and fixable. They won't waste your time hallucinating issues. Note that this is different from "being wrong". A script can be wrong, and you can spend time debugging it, but that time is not wasted. You can fix the script, remove wrong rules, whatever.
Also, nothing exists in a vacuum. Script-generated review comments were never a widespread societal issue, LLMs unfeterred content generation is.
simias · 2026-10-27 11:39:53 UTC
It's a bit insane how it seems that "deterministic and reproducible", what I would consider to be a cornerstone of good software development, has apparently become an afterthought in the age of LLMs. All that for the sake of perceived convenience.
Not even convenience at times, I think that for some the amount of technical-sounding text LLMs can output on demand is comforting. It feels serious. It feels like it's doing something important.
Meanwhile the rest of us have to deal with these low-signal/high-noise walls of text in commit messages and PR comments. Or ask another LLM to summarize them, I suppose. "Doing and undoing, it's all work", as we say in French.
jawilson2 · 2026-10-27 14:57:35 UTC
Devil's advocating here... is human code deterministic and reproducible? If I am assigned some non-trivial task, and write it twice with a week in-between, the code is not going to be identical, though it will meet the requirements. Or, if you give 3 different senior devs the same task on the same day, they will all be different, but still meet the requirements. What is different about asking an LLM to write code?
bigstrat2003 · 2026-10-27 16:13:54 UTC
Well, for one thing, a human actually understands what he's doing. He isn't just generating something plausible based on statistics. As a result he makes fewer mistakes, can catch his own mistakes, and can learn from those mistakes.
pantalaimon · 2026-10-27 17:00:43 UTC
> a human actually understands what he's doing. He isn't just generating something plausible based on statistics.
What makes you so sure that this is not the same thing?
Grombobulous · 2026-10-27 12:33:02 UTC
Is the hallucination or “being wrong” the only major problem? Because if that’s the case I can’t imagine that problem will still be around in ~1-5 years.
I think it’s actually very similar to a script. It takes an input and produces an output. Certainly it has more variation than a traditional script, but it's generally going to return a very similar answer when the same input is given repeatedly, like when you ask an employee at a restaurant where the bathroom is. The verbiage may not be the same every time but the answer is always the same.
cromka · 2026-10-27 11:20:11 UTC
> I'm not anti-ai but I am firmly anti-make-people-read-llm-output-I-didn't-read-myself, and this breaks this.
This here might hit the nail on the head exactly in terms of what the biggest FOSSes gripe with LLMs with might be: not the fact people use it but rather how it hinders the communication.
If you use it to help the write code, but you explain things in your own words or, at very least, redact and trim LLMs output, it's fine.
If you, however, just drop a wall of text at someone and expect they act on its b/s, this is lazy to the point it borders with microagression.
You wouldn't expect a distinguished professor explain basic maths to a student, and we shouldn't expect actual developers address Copilot's stupid, unredacted hallucinations or otherwise unrelated remarks.
This works for patches submission, too: you should be outright banned for submitting stuff you didn't bother to double check and properly document.
close04 · 2026-10-27 12:00:03 UTC
That is one of the biggest gripes for almost anyone I ever met who deals with AI generated content. Nobody really cares that someone else's work was generated with AI as much as they care that it's slop and they need to put in extra effort to work around the author's laziness.
Lazy or incompetent colleagues always forced people around them to overcompensate and correct. AI increased their firepower and blast radius.
bombcar · 2026-10-27 15:08:00 UTC
But you just consume my slop with your Slopmaster 9000™, right?
The end-game for Anthropic is each PR to be 100k+ tokens but nobody cares because they burn 100k+ tokens reviewing and responding.
pluc · 2026-10-27 12:07:11 UTC
If you could explain what you're doing why would you use LLMs to do it in the fist place? For speed over control? Explaining means you're effectively able to do the modification yourself, and I'm fairly certain that's a low percentage of all AI contributions.
phoghed · 2026-10-27 12:24:07 UTC
Same reason I can explain how to clean my house but I pay someone to do it? I want the end result, I don’t want to do the whole process myself.
lifeisloving · 2026-10-27 12:40:45 UTC
No, most AI users who contributor to OS do not actually know what they're doing. Its a problem. And no amount of analogies about house keepers will change that.
bombcar · 2026-10-27 15:12:18 UTC
That's the problem. We have a bunch of people who know how to do the work arguing about using LLMs to do it and assuming that all the submissions/reviews/etc are being done by people like them who know how to do the work.
Once you're drinking from the firehose you realize just how many idiots (well meaning or not) are out there, and how empowered they are now.
unrented7977 · 2026-10-27 16:04:05 UTC
Maybe we should all respond to LLM text walls with aggressively doubled LLM text walls, expanding geometrically until the originating party either gives up or learns to communicate with their own words.
gspr · 2026-10-27 11:55:53 UTC
> They also have a reward system that punishes people for ignoring the llm review (by pushing their submission lower in the queue).
This is what I fear will push its way into every part of society soon, too.
"We're not forcing you to do your banking through our AI interface instead of a normal one, you're free to opt out, but multiple functions will be missing and we'll charge you a monthly fee."
"You don't have to have an LLM grade your exam, but beware that the human grader will be overworked and grumpy for the extra task."
drdexebtjl · 2026-10-27 13:13:06 UTC
This is happening to job interviews right now.
If you’ve applied to jobs recently, you must have seen the companies that transcribe and summarize interviews using AI (and probably share data across their clients in an indirect, technically legal way).
You can opt out, of course, but do you really want to be the one candidate for which the interviewer needs to take notes manually? And they just spent a lot of money on this hiring AI tool, will they hire the one candidate for which they couldn’t use it?
gspr · 2026-10-27 13:28:32 UTC
So I've heard. I also know that several doctor's offices here use AI transcription. I'm sure I have a legal right to demand that they be switched off (I'm worried about privacy and accuracy here), but then I've got to be an asshole and increase the workload on my well-meaning doctor who probably had no say in any of this.
This is gonna get so much worse. It's like the smartphone revolution on steroids – and we've seemingly learned nothing! We're in dire straits.
wccrawford · 2026-10-27 14:06:23 UTC
I went through a lot of medical stuff a few months ago, and every single doctor asked me if they could use the device for that instead of typing it into the computer themselves. One of them explained that it allowed them to focus their whole attention on me, instead of partially on the computer for the inputting of the data while we talked.
I'm surprised that your doctors weren't asking, and instead were just assuming.
gspr · 2026-10-27 14:29:05 UTC
I don't know if my doctor uses these systems. I don't go to my doctor very often. I've heard that many do, and I'm speculating.
jasomill · 2026-10-27 15:29:45 UTC
I don't believe they merely assume. If they don't ask in person, you've probably already opted in through patient check-in forms or the UI of a telehealth app.
solyin · 2026-10-27 13:58:47 UTC
> transcribe and summarize interviews using AI
That's cute. Yo, people be having full on AI-led interviews! That is to say, there's the human interviewee, and an AI, testing the human. Asking them questions and having a conversation that is definitely recorded.
throwawayffffas · 2026-10-27 12:53:59 UTC
Eh, you read compiler and linter output don't you? Nobody read that before you did.
Sure LLMs are not nearly as rigorous about their output but that doesn't mean it's entirely not useful.
At a minimum you can have your LLM summarize the other LLMs output and go from there.
johannes1234321 · 2026-10-27 13:28:57 UTC
Compiler output is somewhat deterministic and there is a logical path (even if caused by a compiler bug) and not dealing with it will definitely(!) lead to a broken product. Also at least I rarely read the full output, but I have an intuition, as the structure is consistent, what to look for in the output.
LLM output may or may not be relevant, may or may not tell me something new and by the way it is structure is often hard to quickly oarse for the relevant piece in a wall of text.
pantalaimon · 2026-10-27 17:26:18 UTC
If I ask Claude to 'Review this PR' it will usually generate a very useful output and find bugs I wouldn't have spotted.
It does not have to be a wall of slop, maybe you only used LLMs 2 years ago where this would have been all you'd get.
vitorsr · 2026-10-27 13:36:01 UTC
Maybe it's an unpopular opinion, but a concise report from one of today's frontier models is not something I would be offended by, even if nobody read it beforehand. I guess the fact that there is a nontrivial cost to it also eases the existential angst of neither being deterministic (Lintian) nor having been produced by a person.
bayindirh · 2026-10-27 13:48:49 UTC
Disclosure: I'm not a Debian Developer, yet I have worked on Debian derivatives and I'm somewhat familiar with internals of Debian from technical and human perspectives.
While I'm perfectly on the same page with the critics of the LLM generated summaries, I'm a bit torn on the issue.
For anyone not familiar with Debian ecosystem, there's a helper tool called Lintian. Lintian does all kinds of checks from package formatting, licenses to dependency management (which can detect both unnecessary yet included and necessary but missing dependencies via static analysis of your tool if necessary).
It's expected that your package to be "Lintian clean" to be able to included in Debian with minimum friction and wait. Even though it's Lintian clean, your package also checked by teams which are severely understaffed, and composed solely of voluntary people.
NEW queue also receives tremendous influx of packages (which can be considered a torrent, even). Since noon, there are 10 more ITP (intention to package) bugs open in my inbox.
As a result, at one hand, these people need every help they can get because getting new trainees and getting them to learn the intricacies is hard, on the other hand LLM generated text rightfully generate negative reactions from people. However, giving developers an automated check of what looks right and what looks wrong is helpful, in my perspective.
I wonder what people would have said it was a deterministic tool generating these reports and sending people these summaries as a "hey consider this feedback until a human has time to see your package".
This is a hard question and we need to think before impulsive reactions. I even don't have all or any of the answers, but at least I have some insight why this is there.
monegator · 2026-10-27 14:12:09 UTC
one of the points is that LLMs are NOT deterministic so you can't run them yourself a thousand times and expect the same output.
bayindirh · 2026-10-27 14:16:31 UTC
That's true, but if the LLM is calling a tool to provide the base information to extrapolate, the results will not change even if the sentence structure is never the same.
Debian has an extensive toolchain to check things. Having a set of deterministic tools which feed LLM with data can be beneficial, though.
Again, I'm not defending the LLM here, just want to show how hard the problem is to begin with.
khaelenmore · 2026-10-27 16:12:50 UTC
And LLM "summarizes" the tool output as a completely unrelated and at times self-contradicting text.
If there are deterministic tools involved - just return their output as is. It's usually quite readable if needs a bit of adjusting
pantalaimon · 2026-10-27 17:23:49 UTC
When was the last time you used a LLM?
This is not an issue with current models.
cadamsdotcom · 2026-10-27 10:54:08 UTC
This is a strongly phrased message but the problem could be solved with an opt-out.
Even something as simple as writing the phrase "No LLMs please" somewhere in the upload, that if present, prevents LLMs being triggered. Nothing elaborate and it preserves the default-on behavior the team seems to want.
philipwhiuk · 2026-10-27 12:21:01 UTC
> the problem could be solved with an opt-out.
The problem would be better solved with an opt-in.
silveraxe93 · 2026-10-27 14:14:06 UTC
Yes, the point of the article is that he'll stop uploading _unless_ there's an opt out.
> I will not upload anything to NEW unless there is a way for me to opt out of
this. I would prefer if the default was "no LLM review" and that developers
have to explicitly opt-in to receive an auto-generated review.
baumschubser · 2026-10-27 10:57:44 UTC
i wozld not care for most other actors even in the FOSS universe to swallow their pride and put LLM into their workflows.
However, Debian proved in countless occasions before that they can and will go the route that is right, not the one that is easy, even if it makes everything slow or cumbersome and I think thats a good thing and the reason why so many others build their stuff in top of it.
mono442 · 2026-10-27 11:05:03 UTC
There's no point volunteering free time just to deal with slop.
boredhedgehog · 2026-10-27 11:13:42 UTC
So how often is this LLM wrong?
toenail · 2026-10-27 11:32:45 UTC
Developers shouldn't have to waste their time to figure this out.
iugtmkbdfil834 · 2026-10-27 11:33:47 UTC
I think this the wrong question to ask. The original complaint suggests that llm setup used is largely unknown in terms of how it organized. For all you know, it may be splitting and sending different requests to unpaid gemini, local 8b qwen or paid claude/astra. None of it is known, listed or noted as expected to be somewhat stable.
The question is not even about the setup in terms of hardware/software. The question is about organizational structuring of its processes.
Grombobulous · 2026-10-27 13:46:33 UTC
It sounds like the author doesn't know, but I suspect that the DFSG knows exactly how it is setup.
gilrain · 2026-10-27 14:41:44 UTC
Every time its use insults (in aggregate) invaluable human contributors. It doesn't even matter if they're philosophically "correct" or not: it's about respect.
vconnor · 2026-10-27 11:23:53 UTC
Elsewhere in the email thread:
Picking at random a successful and unsuccessful LLM-generated
review from the dashboard, the successful review consists of 734 words
(~2.5 printed pages) and the unsuccessful review weighs in at a
whopping 1,535 words (~5 printed pages). I haven't tried to fully read
either report, nor spend brain cycles checking if the LLM output is
legitimate or not. I just want to emphasize the many orders of
magnitude change in review size that's being generated, and that
uploaders may need to spend time trying to review and understand in
this new process.
A bit rude to expect people to swallow 5 pages of slop to figure out why review failed.
EDIT: made my comment more constructive, but originally asked if any distro has adopted a hard anti-AI stance.
none_to_remain · 2026-10-27 11:26:20 UTC
No Linux distro can have a hard anti-AI stance
vconnor · 2026-10-27 11:29:13 UTC
A distro is not responsible for its packages, but it can be responsible over its processes.
Linux itself is in any case tainted, but some of us care to contain the plague however possible.
account42 · 2026-10-27 11:38:03 UTC
Of course it can. It would mean not including some software or hard-forking it from pre-AI versions which would be difficult. But it can be done and I would seriously consider using such a distro if it was available.
DaSHacka · 2026-10-27 16:57:00 UTC
Well, the BSDs are right there.
bigfishrunning · 2026-10-27 12:04:39 UTC
Why not? Linux distributions contain tons of packages that are not the Linux kernel, and just because AI patches have been accepted in one place doesn't mean they'll be accepted elsewhere.
Grombobulous · 2026-10-27 13:48:53 UTC
Because Linux distributions must contain the Linux kernel, which is built with the assistance of AI tooling.
It does have very strict standards, but AI is explicitly allowed to be used in the aid of development of the Linux kernel.
You can't be hard-no-AI as a Linux distribution because it would be like having a vegan car company where the seat supplier used leather. You'd have to fork the Linux kernel from the last version that was developed before LLMs were adopted, or write your own kernel.
BSDobelix · 2026-10-27 11:32:17 UTC
>>Any distro that have adopted a hard anti-AI stance?
Yes all the BSD's, and i think with the best explanation, since one cannot proof what license the llm's output has, one cannot legally implement the output under the BSD-License.
But what about Linux? They just dont care, they have the money, corporate backing and with that, all the power.
account42 · 2026-10-27 12:05:56 UTC
Unfortunately, as we can see here isn't not just the corporate projects that are susceptible to people ignoring morals in the name of pRoDuCtIvIty. It's really sad to see Debian of all distros, who where one of the major sources of what it means for software to be free, to go down this road and so quickly.
BSDobelix · 2026-10-27 12:57:03 UTC
Maybe this is the time to revive HURD?
znpy · 2026-10-27 12:02:16 UTC
I’d rather have a detailed explanation of why my contributions are being rejected rather than the complete opposite which is often something along the lines of “no”
frou_dh · 2026-10-27 11:29:29 UTC
I think some developers are kinda dense or something, and don't understand that shoving AI content in people's faces is simply rude.
In recent versions of iTerm2 for example, when using the plain old menubar and hovering over some items, panels appear showing awful sloppy AI images. WTF? https://files.catbox.moe/5ivkhw.png - That makes me question everything... Was the developer always tasteless? Or did they get a taste bypass?
account42 · 2026-10-27 11:36:11 UTC
Oh god they have invented useless hero images for tooltips. The images being slop is just the cherry on top.
tjpnz · 2026-10-27 11:42:46 UTC
Oh jeez, that's atrocious. I'm starting to reconsider some of the things I used to say about Corporate Memphis.
ninkendo · 2026-10-27 11:56:46 UTC
No, it is always right to criticize Corporate Memphis. Keep up the fight.
frou_dh · 2026-10-27 14:05:08 UTC
I dunno, tjpnz might be on to something. Would a nice smooth blue leg stepping towards the camera with super exaggerated perspective really be so bad right about now?
alex_suzuki · 2026-10-27 11:48:12 UTC
WTH…? I use iTerm2 and wasn’t aware of this. What an incredibly bad call.
meindnoch · 2026-10-27 11:49:04 UTC
Yes, they were always tasteless. One job of a platform SDK is to enable creating tasteful things that mesh nicely with the platform idioms, and make it hard to create tasteless things.
With AI, creating the tasteless thing became easy.
If you give people the ability to select a font, they will select Comic Sans.
rf15 · 2026-10-27 12:04:39 UTC
... in pink on yellow too, if you add a color picker.
wyclif · 2026-10-27 12:22:12 UTC
Seeing this makes it hard to believe that I actually used iTerm2 for a few years back when I was still on macOS.
mrob · 2026-10-27 13:53:13 UTC
Giving developers the ability to impose Comic Sans on the users is bad. Giving users the ability to make their own tasteless choices is good. It's my computer and I'll make it ugly if I want to. Windows 3.1 included the famous "Hot Dog Stand" theme. The loss of this freedom only encourages user-hostile practices.
superdisk · 2026-10-27 12:54:47 UTC
I think iTerm 2 jumped the shark when the author started implementing a homemade web browser into it.
That AI image is pretty baffling though. Setting aside the pointlessness of displaying an image in the context menu, it even looks like it was made with ancient stable diffusion from 2024 or something, modern image generators are "better" than that. I wonder when it was put in.
Comments
There is no alternative to Debian for me. I’m ready to justy throw out all my computers and become a goat herder at this point.
PS I’ll remove memory and storage from said computers to pay for land and goats before I chuck them out of my window.
https://marc.info/?l=openbsd-tech&m=177411620801633
Lesson for us all is that if we do not own the thing/project/etc, things can suddenly change without our decision, and all time we put into it might become a waste-of-effort retroactively
I contribute to open source software because it improves stuff I use and/or care about.
It seems too pessimistic, IMO, to abstain because things could potentially change to the worse, sometime in the future.
Patience and persistence helps, but it can be daunting IMO
I replied as I did, because lukeify implied that lack of ownership is
> one of the primary reasons I can never get interested in contributing to open source
Open source is open source. publishing code.
it does not mean
- you need to respond to peoples PRS
- it does not mean you need a github
- it does not mean you need to provide ANYTHING to the user. No warranty.
- you do not need a mailing list
- you do not need governance
Do what you want, have fun, if you have peers or other people interested in your area, feel free to collaborate in the same way you would with pickup basketball at the park, play, have fun, and leave when you're done.
https://www.debian.org/devel/constitution
It sounds like in this case someone is not following an agreed-upon policy. And of course, even with governance like this one, the rest of the maintainers can move in a direction you don't agree with.
They also have a reward system that punishes people for ignoring the llm review (by pushing their submission lower in the queue).
But to upload to debian new's queue you already need to have pre-approval, it's a different situation from e.g. a public git repository that anyone can push to.
Why would you use an LLM instead of a fixed message?
Another situation where I would use llms to generate messages is if we've mutually agreed to do that. For example automated review messages at work, if we've agreed that it would be a part of our workflow.
A good rule of thumb might be unsolicited vs solicited llm messages.
If you mutually agreed of course it's fine
For example, mine is a technical writing skill based on the old book "The Art of Scientific Writing": Front load information as much as possible - the first paragraph of a document should be a summary, the first sentence of a paragraph should summarize. Short sentences building on each other. Always show evidence.
Someone else spent quite a while optimizing a standardized PR description template, with a section for intent, expected changes, risk, rollout coordination and such.
That made the slop a lot more tolerable.
A script is reproducible, and fixable. They won't waste your time hallucinating issues. Note that this is different from "being wrong". A script can be wrong, and you can spend time debugging it, but that time is not wasted. You can fix the script, remove wrong rules, whatever.
Also, nothing exists in a vacuum. Script-generated review comments were never a widespread societal issue, LLMs unfeterred content generation is.
Not even convenience at times, I think that for some the amount of technical-sounding text LLMs can output on demand is comforting. It feels serious. It feels like it's doing something important.
Meanwhile the rest of us have to deal with these low-signal/high-noise walls of text in commit messages and PR comments. Or ask another LLM to summarize them, I suppose. "Doing and undoing, it's all work", as we say in French.
What makes you so sure that this is not the same thing?
I think it’s actually very similar to a script. It takes an input and produces an output. Certainly it has more variation than a traditional script, but it's generally going to return a very similar answer when the same input is given repeatedly, like when you ask an employee at a restaurant where the bathroom is. The verbiage may not be the same every time but the answer is always the same.
This here might hit the nail on the head exactly in terms of what the biggest FOSSes gripe with LLMs with might be: not the fact people use it but rather how it hinders the communication.
If you use it to help the write code, but you explain things in your own words or, at very least, redact and trim LLMs output, it's fine.
If you, however, just drop a wall of text at someone and expect they act on its b/s, this is lazy to the point it borders with microagression.
You wouldn't expect a distinguished professor explain basic maths to a student, and we shouldn't expect actual developers address Copilot's stupid, unredacted hallucinations or otherwise unrelated remarks.
This works for patches submission, too: you should be outright banned for submitting stuff you didn't bother to double check and properly document.
Lazy or incompetent colleagues always forced people around them to overcompensate and correct. AI increased their firepower and blast radius.
The end-game for Anthropic is each PR to be 100k+ tokens but nobody cares because they burn 100k+ tokens reviewing and responding.
Once you're drinking from the firehose you realize just how many idiots (well meaning or not) are out there, and how empowered they are now.
This is what I fear will push its way into every part of society soon, too.
"We're not forcing you to do your banking through our AI interface instead of a normal one, you're free to opt out, but multiple functions will be missing and we'll charge you a monthly fee."
"You don't have to have an LLM grade your exam, but beware that the human grader will be overworked and grumpy for the extra task."
If you’ve applied to jobs recently, you must have seen the companies that transcribe and summarize interviews using AI (and probably share data across their clients in an indirect, technically legal way).
You can opt out, of course, but do you really want to be the one candidate for which the interviewer needs to take notes manually? And they just spent a lot of money on this hiring AI tool, will they hire the one candidate for which they couldn’t use it?
This is gonna get so much worse. It's like the smartphone revolution on steroids – and we've seemingly learned nothing! We're in dire straits.
I'm surprised that your doctors weren't asking, and instead were just assuming.
That's cute. Yo, people be having full on AI-led interviews! That is to say, there's the human interviewee, and an AI, testing the human. Asking them questions and having a conversation that is definitely recorded.
Sure LLMs are not nearly as rigorous about their output but that doesn't mean it's entirely not useful.
At a minimum you can have your LLM summarize the other LLMs output and go from there.
LLM output may or may not be relevant, may or may not tell me something new and by the way it is structure is often hard to quickly oarse for the relevant piece in a wall of text.
It does not have to be a wall of slop, maybe you only used LLMs 2 years ago where this would have been all you'd get.
While I'm perfectly on the same page with the critics of the LLM generated summaries, I'm a bit torn on the issue.
For anyone not familiar with Debian ecosystem, there's a helper tool called Lintian. Lintian does all kinds of checks from package formatting, licenses to dependency management (which can detect both unnecessary yet included and necessary but missing dependencies via static analysis of your tool if necessary).
It's expected that your package to be "Lintian clean" to be able to included in Debian with minimum friction and wait. Even though it's Lintian clean, your package also checked by teams which are severely understaffed, and composed solely of voluntary people.
NEW queue also receives tremendous influx of packages (which can be considered a torrent, even). Since noon, there are 10 more ITP (intention to package) bugs open in my inbox.
As a result, at one hand, these people need every help they can get because getting new trainees and getting them to learn the intricacies is hard, on the other hand LLM generated text rightfully generate negative reactions from people. However, giving developers an automated check of what looks right and what looks wrong is helpful, in my perspective.
I wonder what people would have said it was a deterministic tool generating these reports and sending people these summaries as a "hey consider this feedback until a human has time to see your package".
This is a hard question and we need to think before impulsive reactions. I even don't have all or any of the answers, but at least I have some insight why this is there.
Debian has an extensive toolchain to check things. Having a set of deterministic tools which feed LLM with data can be beneficial, though.
Again, I'm not defending the LLM here, just want to show how hard the problem is to begin with.
If there are deterministic tools involved - just return their output as is. It's usually quite readable if needs a bit of adjusting
Even something as simple as writing the phrase "No LLMs please" somewhere in the upload, that if present, prevents LLMs being triggered. Nothing elaborate and it preserves the default-on behavior the team seems to want.
The problem would be better solved with an opt-in.
> I will not upload anything to NEW unless there is a way for me to opt out of this. I would prefer if the default was "no LLM review" and that developers have to explicitly opt-in to receive an auto-generated review.
The question is not even about the setup in terms of hardware/software. The question is about organizational structuring of its processes.
Picking at random a successful and unsuccessful LLM-generated review from the dashboard, the successful review consists of 734 words (~2.5 printed pages) and the unsuccessful review weighs in at a whopping 1,535 words (~5 printed pages). I haven't tried to fully read either report, nor spend brain cycles checking if the LLM output is legitimate or not. I just want to emphasize the many orders of magnitude change in review size that's being generated, and that uploaders may need to spend time trying to review and understand in this new process.
A bit rude to expect people to swallow 5 pages of slop to figure out why review failed.
EDIT: made my comment more constructive, but originally asked if any distro has adopted a hard anti-AI stance.
Linux itself is in any case tainted, but some of us care to contain the plague however possible.
It does have very strict standards, but AI is explicitly allowed to be used in the aid of development of the Linux kernel.
You can't be hard-no-AI as a Linux distribution because it would be like having a vegan car company where the seat supplier used leather. You'd have to fork the Linux kernel from the last version that was developed before LLMs were adopted, or write your own kernel.
Yes all the BSD's, and i think with the best explanation, since one cannot proof what license the llm's output has, one cannot legally implement the output under the BSD-License.
But what about Linux? They just dont care, they have the money, corporate backing and with that, all the power.
In recent versions of iTerm2 for example, when using the plain old menubar and hovering over some items, panels appear showing awful sloppy AI images. WTF? https://files.catbox.moe/5ivkhw.png - That makes me question everything... Was the developer always tasteless? Or did they get a taste bypass?
With AI, creating the tasteless thing became easy.
If you give people the ability to select a font, they will select Comic Sans.
That AI image is pretty baffling though. Setting aside the pointlessness of displaying an image in the context menu, it even looks like it was made with ancient stable diffusion from 2024 or something, modern image generators are "better" than that. I wonder when it was put in.