Removing my nix flakes vs guix post
86 points by untrusem
86 points by untrusem
I have known the author personally for quite a while, I am sad they had to go through this.
Frankly, this is the fallout of the constant "post is clearly LLM" comments everywhere and the tagging of everything with vibecoding for the most minute reasons. Because all it takes is one person to write the accusation and there is almost no defense to proving you didn't use an LLM, especially if you also post opinions about LLMs or have LLM related posts in your blog. Guilty of spam until proven innocent.
I'm pretty virulently opposed to Gen AI, but I agree the constant accusations are getting tedious. I'm old and not good at recognizing the supposed tells, but I'm not convinced the people determined to jump on every single thing are actually as good at it as they think they are, either. LLMs mimic a certain style often used by humans in marketing and other business oriented writing before the AI age, and lots of other humans will have subconsciously picked that up, for better worse.
I absolutely want to know if code is created with AI (for reliability reasons), and with images it's usually pretty obvious (and again, I want to know if not, for reality-testing reasons) .. but text you can skim and make a judgement about the SNR and quality pretty easily. There's less reason for hair-trigger reflexes there.
Exactly. And this is why I'm getting more and more reluctant to contribute to this site by submitting my own stories, which in the past have often been well-received.
I was already accused once to have written an article with an LLM (which I didn't), and as my most recent articles are about AI, the most obvious conclusion is that they are written with AI, and whatever follows next is likely tainted. Right?!
You'd just say "oh it's just a tag", but it's more than that: it's the broken respect of a community which once felt very balanced, but seems to be broken these days. Apparently I'm not the only once thinking this based on the high number of upvotes in this thread...
The problem is that everybody wants their cheap "This article likely won't suck"-proxy back. Unfortunately, it isn't coming back--at least certainly not until the money drains out of the AI bubble.
Articles used to be "good" or "bad" or "factual" or "erroneous"--all of which had quick and obvious tells as you read the article. The current issue is that the existence of LLMs produces a third category of negation--"Not obviously wrong but insipid and a waste of my time."--that takes too much effort to sort out. This is the "Not even wrong" of Wolfgang Pauli but supercharged by computer heat and VC money.
Even worse, I can't even trust that the "This is LLM" crowd--ARE ACTUAL HUMANS! It's turtles all the way down.
A friend of mine sent me this article before I saw it on lobsters.
I remember reading it and having the urge of accusing it of being LLM slop, but for whatever reason I decided to push through it and articulate a more specific critique. I ended up writing "kinda feels like the author didn't quite internalize why flakes were necessary, but it is cool" (and some more discussion followed about how flakes do things and why).
In retrospect, I'm glad I didn't reply with a lazy spam accusation. It's a short-circuit that would have prevented me from taking in more of the content (which was interesting even if not comprehensive!!), and it would have prevented the conversation we had after.
My personal takeaway is that I'm going to try to reduce the weight of tools I suspect people used for research. If it's outright spam and effectively a DoS of attention, it's fine to disregard regardless if it was done with flapping meat or LLMs. But if it's something that is not quite fitting my mental model, I hope that I spend more time trying to tease out a more constructive critique rather than focusing on their tool use.
The increasing urge of labeling things as LLM slop says more about us, and our reaction to the changes in our society, than it does about the author whose work we might be mistaking.
If you don't think LLMs are a bad thing, why is it a problem? And if you do, you can see the damage and erosion of trust that allowing LLM output into human spaces causes. I don't want things to be like this, but it's the future we raced to build.
We broke things. No amount of wishful thinking will fix that. The world is now going to be worse forever. The only thing that can help rolling things back even a little is setting up spaces where LLM lovers stop showing up.
People have come by with baseball bats and smashed up the place. Damage has been done. We can't exactly go back. And it's frankly offensive that people are saying "people are trying to see I'm holding a baseball bat, things feel so unwelcoming...". It turns out that more often than not, the people complaining are the ones holding a baseball bat.
Yes, this harms people. No shit. Why do you think people are so strongly against GenAI? Because there's no way to use it without causing harm.
So your solution, when somebody who isn't holding a baseball bat and didn't smash anything up gets accused of smashing things with a baseball bat, is to say "Welp, we broke things and people are getting hurt, sucks to be you?"
No, the solution is to apologize and move on. The best we can do is to make LLMs clearly disallowed, which will reduce the number of times people have to ask.
But we built a shitty future, and now there's no way to get away from this. If I had a magic wand, I might be able to do something to put together a world where things weren't shitty.
Do you have a perfect solution, or are you just advocating more tolerance of antisocial behavior?
I'm advocating not making things worse by throwing around blind accusations on every single article that gets posted. The way to build a human community that emphasizes our shared humanity is to treat other people in the community with respect and dignity.
First, that doesn't solve a damn thing, and making it clear that people are potentially about to click on a denial of service attack on their time is a valuable service, in the absence of a clear ban.
Second, are they blind? So far it seems like they've had a pretty solid hit rate. Do you have a list of accusations where it turned out that LLMs were not used? I can think of maybe one or two.
Would I prefer a world where this necessary, and we had a 100% accurate way of zapping only LLM posts? Of course, but quietly accepting slop posts doesn't move us in that direction.
I do agree though, I don't want to see people posting on threads and tagging. I want the moderators to moderate. Bans should be handed out.
So, yes. This situation sucks. I don't see many viable alternatives for coping with this future we've built that isn't just capitulation.
Second, are they blind? So far it seems like they've had a pretty solid hit rate. Do you have a list of accusations where it turned out that LLMs were not used? I can think of maybe one or two.
You are literally commenting on an article where this happened. I've had it happen to me. Other folks in this same thread have mentioned having it happen to them. I'm sure there are others who got incorrectly accused of being a robot and then just shut up and went away because they were too tired/scared/intimidated to stand up for themselves.
You are literally commenting on an article where this happened.
Yes. That's one of the ones I was thinking of.
Now, what do you propose to do that doesn't have any collateral damage? Keep in mind that I'm aware of people who have already left this site (and the computer industry entirely, for that matter) because of the tsunami of AI; today they make in a day what they used to make in an hour, and still claim to be happier for it. They're still writing code in their spare time, but they've entirely disengaged with commercial software development. Ditto for folks in other industries who are being impacted by this. If you chose to capitulate, you hurt people like them.You don't get to pretend that shutting up about the people AI is hurting somehow leads to a better situation.
What does empathy for the people that have been pushed out of software by AI look like? Think hard about the harm your suggested capitulation will do.
I don't know how to prevent the voices of humans from being drowned out; I don't see a better option than ensuring that we have spaces that take a strong stance that about being exclusively for humans, and defend that stance strongly. Otherwise, we lose the spaces where humans can have a voice at all, without getting drowned out in generated content. Maybe that doesn't matter, and human writing is something to be tossed into the trashbin of history.
Encouraging baseless accusations against genuine human writing is only going to accelerate that process. Why bother writing, if some mob is just going to scarlet-letter your work because it “smelled LLM-y” to them?
Besides which, as LLMs advance, and get better at imitating humans, these accusations are going to get more baseless. It’s a losing proposition.
The accusations are fundamentally not baseless, they're based on intuition. I don't share the assumption that LLMs will get better at imitating humans. The fight for the voice of real people over robots is worth potentially losing, sorry.
Again, waiting for any mention of a proposed solution. It sounds more and more like the prevailing opinion is that it's time to give up on the Internet.
Why bother putting effort into writing, if it's just going to be tossed on the pile of slop, where most articles are generated?
How many people do you think have already given up? Have you talked to young folks about the future we are leaving for them? Taking a stand and saying "we are making space for humans" is about as welcoming as it gets for people. Giving into nihilism is alienating.
The only proposed solution here, which I’m cautioning against, is banning humans from the community because of somebody’s “intuition”. The OP was not “welcomed”, they were dismissed. We’ve always had people being rejected because of their content, but now we’re going to do it because of their style?
Maybe folks around here have a 100% accurate “intuition” (some of them sure think they do), but in the broader world I’ve seen people accused of being LLMs basically because they have a large vocabulary and write in complete sentences — not to mention knowing how to use an em-dash. That is not acceptable to me.
Yes, that's the problem; nobody has any viable alternatives. If you want to propose something else, I'm all ears, because I'm very aware that there are difficulties when putting together a space that is not hostile to people.
Today, we are building online spaces that are hostile to people, and I'm rapidly reaching the point where I don't want to participate at all. From talking to friends, it seems I'm not alone.
The current situation is driving away human writers.
I don't agree that I have to propose an alternative solution to be allowed to point out that the currently proposed solution isn't a solution — and in my view is actually making the problem worse.
The internet is not a community. This is a community. And from my point of view parts of this community are advocating being hostile to people by labeling their writing as non-human, and the willingness to throw out the humans with the AI bathwater makes me wonder if I want to participate in this community.
All online communities are on the internet, and if your argument applies to one, it applies to all of them. Essentially, you are saying that online communities should accept AI output, either explicitly or implicitly.
I've got zero interest in participating in a space like that, so if you get your community, you are choosing to push people me out. That's a valid choice you can make. but it's an active choice on who should leave the community.
Given that AI is currently less popular than ICE in the USA, that's an awfully big portion of the population (though, I'm not sure how well that reflects the people on here).
I totally agree AI content is poisonous and unacceptable. What I don’t agree with is banning/flagging alleged AI content based on nothing but (ironically) vibes. Because you absolutely will ban/flag real humans, as has been demonstrated.
Maybe we need a separate voting dimension of “perceived humanity” and you can turn the dial to the threshold you want, based on your opinion of the crowd’s AI detector reliability.
(looks like you added the bit about a shadowbanning vote. I think it's a good idea. Shadowbanning has its downsides too, of course. I also edited in in my thoughts.).
So, when someone posts AI content, what is the action taken to protect the community against this "poisonous and unacceptable" content?
"Unacceptable", to me, means not accepting it. I'd like to know what action that implies for you.
For me, what I'd like is a clear statement of values from the site. The rest is policy which can change over time.
I think that a good medium-long term state is a hidden "non-human" flag that prompts moderator action, as well as a suite of tools that help them make judgements on removing content. With clear values and best effort enforcement, I'd love to mark debates on the humanity of content off-topic, and delegate case by case calls. I am not thrilled by the moderation team not making any statements on this site's values.
are you just advocating more tolerance of antisocial behavior?
From where I stand, this is what you're doing. You're the one asking why someone who doesn't think LLMs are bad would not want to be accused of having not written their own prose, to try and drown out people who are saying that they're uncomfortable accepting false accusations.
If you genuinely believe LLMs are just a tool, then saying someone is using an LLM is no worse than saying someone is using emacs.
To have a problem with someone saying someone is using an LLM, you have to have empathy for the anti-LLM stance, and the position that it's not just a tool that you can use without impact. At that point, the question shifts to how to minimize the impact; I personally don't see a way to route around the damage that LLMs have done without any downside. That ship has sailed.
Even a moment's thought would've revealed that someone could consistently be ok with some uses of LLMs and not others. Asking the question answers it. You didn't even ask.
You can hire people to write things for you. This is often ethical. You can hire a ghostwriter to write a book for you. This is considered less ethical. You can hire someone to write messages to your crush as if they were you. This is generally considered wildly unethical and shameful.
The same goes for using LLMs. Creating code and other artifacts with them is different from writing a blog post that is in your personal name. Or perhaps you don't think they are different--that is your prerogative. But you need the basic theory of mind to not assume that all others think the same as you.
Your tone says you disagree, but the words don't seem to.
I very specifically disagreed--you can think a tool is ok, and be opposed to specific uses of it. In that case, you'll not appreciate being accused of using it in those cases.
But going further, even if you think LLM writing is ok, you may still make personally make a different tradeoff, and resent being dismissed as producing LLM output.
If I'm a great chef or musician or athlete, I don't have to think that people who cook a frozen dinner or don't practice guitar or never do more than casually job are bad people. But I wouldn't take it kindly if someone said "he's just sort of a casual cook/musician/athlete".
Yes, and if you think your use of it is ok, why should you object if someone points out that you used it? I think you typed your response with a keyboard -- what an insult, right?
If you think that it's a problematic statement, you're implicitly implying that there's something wrong with using a keyboard for typing a response. The reason people are hurt by false accusations of using LLMs for writing is because of their beliefs on how LLMs should be used.
They can object for any one of the following reasons:
[0] Why not? Perhaps they value the intellectual exertion more than the outcome. Perhaps they believe LLM output is less polished. Like a frozen dinner, you don't always optimize for quality because the existence of the outcome in minimal time is what matters ("I could've written a shorter letter if I'd taken more time"). Perhaps they want their own voice to show, but acknowledge not everyone has to care about that.
I don't get why this post gets flagged. Wrong accusations of LLM usage are a thing with (and you can read in this post) real consequences on writers' mental health and motivation to write. This is discussion-worthy topic.
I say this as a person with a pretty thick skin who doesn't much care what randos on the internet think about my writing: getting accused of using an LLM to write words, when in fact I did no such thing, is one of the most demoralizing and frustrating experiences I've ever had. Maybe we should stop doing that.
Try some cognitive restructuring. When people indicate that they think I used an LLM, I don't think, "Oh, I'm deeply insulted because I've been compared to a chatbot." Rather, I think, "Oh, they aren't skilled enough with words to discern good writing from bad writing."
I don't love that you're implicitly suggesting to someone who has been substantially wronged that they need to get a thicker skin about it, but on the other hand you are offering specific concrete advice about how to do that, rather than just telling them to figure it out.
I think that what you're suggesting is a good idea because Lobsters does already have some other issues where cognitive restructuring seems to me to be the simplest way to brush them off. One example is the fact that anything you submit will be replied to by a minority of people who didn't read it.
As someone who's benefited from CBT in the past, I can say that this class of advice does feel this way sometimes!
"Yeah that thing that happened that upset you so much and led to these problems? It wouldn't nearly as much, if you thought about it this way instead."
It can also seem dangerously close to the primacy of consciousness fallacy.
"I reject your reality and substitute my own"
But properly understood and applied it's neither, and is at least in my experience very helpful.
It's particularly important to realise that there is an objective reality which you are experiencing; else you wind up somewhere philosophically daft.
there is an objective reality which you are experiencing; else you wind up somewhere philosophically daft.
Platonism isn't daft :(
But yes cognitive restructuring type advice can also result in a nasty trick where the fault lies on the person aggreived. Sometimes it is the world that is wrong. (Mark Fisher talks about this, somewhere, about depression and the economy.) It is wrong that there exists a device that makes writing that wastes everyone's time. It is wrong that this device makes everyone mistrustful of each other.
Platonism isn't daft :(
"Daft" was unkind of me; what I was getting at was that the moment you posit a non-objective reality, you've entirely left the realm of several very useful things including experimental (dis)proof and rigorous logic. I would argue that any argument that posits non-objective reality undoes argumentation itself.
Completely agreed re. cognitive restructuring advice, though. And that caution points back to objective reality: sometimes it is the world that is wrong, and that can only be true if the world really exists.
The person whose blog post inspired the removal updated their post with an interesting statement:
P.P.S. I'm sad to see feelings of the author got hurt by my suspections. In fact, I didn't even care much if their post was hand-crafted or not:
Despite the way that post is written (with LLMs or not)
And definitely didn't care enough about a potential human on the other end of the cable. I appreciate style, art and craft, but it was compeletely out of my focus at the moment of writing (my eyes still can barely focus on close up objects yet, heh).
It's very likely I can understand the feelings and moreover can probably relate. I had a similiar expreience with my very own writing about times of my life in the forest. Somebody shared it on hn and in a few hours it was accused of being AI-generated and drug induced and then flagged. I know how unpleasant it can be. I also know how discouraging it is to get a paper review with devaluing comment in the first sentence.
I'm always sad, when my actions unfairly hurt people's feelings and discourage them. I'm always happy to be wrong about my pessimistic assumptions.
I find it so weird and frankly a tad off putting that there isn't even a single "sorry" in here. Interpreting this purely in a vacuum, the author has experienced similar, has the empathy to understand this hurts a lot, reflects that they did a bad thing by judging the person behind the post shallowly... and then doesn't do the extremely obvious last step to remedy the situation.
As for the site itself, I also experienced the mobile glitch with the site before it was taken down and that was really annoying, but beyond that it was obvious that there was a huge amount of care put into it. Guix is a really cool system, but like any GNU project it's ragtag and finding good documentation can be a bit of a lottery. So posts that put its concepts into parlance more people can understand are a great boon to all present and future users.
That reads exactly like an empathetic apology to me. Is the problem that they didn't write the word "sorry" in it?
It shows some empathy, but there's no commitment to do better in the future, or even acknowledging that it was wrong. Just that the actions had negative consequences and they relate to them (wrong or not).
"I feel bad that what I did made you sad" is not an apology.
And definitely didn't care enough about a potential human on the other end of the cable
This part seems like owning that they did something wrong, and will seek to do better in the future. (enough being the operative word)
At this point I have an aversion to editing my posts at all. The more it looks like the initial ramblings of a madman (how my first draft looks), the least I can get accused of using an LLM to write posts.
You're onto something here, maybe that will be the end of my own writer's blockade if we start considering good writing to be an anti pattern.
I'm sad to see the post taken down - I thought it was a useful and thoughtful comparison of guix and nix. Even more so as it inspired another post which tries to correct some errors. For readers, both are valuable - and the second would not have been written without the first.
I'm sorry there were accusations of LLM use, as I thought the post was well written. I could maybe see that from the critics point of view that the post was a bit "confidently wrong" in places - typical of LLM "research". But also typical of a non-expert trying to navigate baroque internet sources on a subject, and trying to figure out which "facts" are still current.
I think being wrong on the internet is invaluable - as it frequently really does prompt someone else to post a rebuttal.
I hope you keep writing, @coopi.
I think the "this was written by AI" takes are going to become the "the image looks fake, I can tell by the pixels, I've seen a few 'shops in my day" of days gone by.
Note that the Lobsters thread, besides one deleted comment (that I cannot know what it was about), there were 0 comments accusations of LLM authorship. (Mostly there was criticism of the JS issues.)
My reading of the post is that the author was affected by one guy's article. Maybe the guy was influenced by Lobsters, but I would think that is a stretch.
I do agree that LLMs are causing (maybe surfacing) a lot of damage. And a good chunk of that comes from people who oppose LLMs! We need to find a way to not cause harm to individuals.
IMHO, on Lobsters we're no longer supposed to discuss endlessly about whether an LLM wrote an article or a piece of code. You're supposed to flag "content that is created without meaningful human authorship" and move on. I don't think that's ideal, but flagging is always going to be objective, and I'm not sure that you can do much about that.
However, I don't think Lobsters members are really 100% aware of that yet. I feel there must be more effort in this direction.
The other is that there are reasonable complaints about the vibecoding tag. Honestly, I don't believe in filtering (precisely because it's "dangerous"); I really like how tags like "video" help me decide whether I click on a link, go on to read comments, or ignore the story, and I think that's how tags should be used. But nonetheless, I feel "we" need to do some more thinking on LLMs and Lobsters, and prioritize being good to others.
Some days I could just fucking scream at the way we all can't figure out how to help each other
As someone curious about nix and guix, the original article was genuinely one of the most informative and enjoyable articles I have read in recent weeks on lobsters and I am sad that the author felt he had to take it down. It was a fantastic introduction/overview.
This is sad to see. There's always the fear a blog post will explode in your face, and why I'm impressed anyone ever posts anything, or submits their own content to any aggregator. There's a lot of folks to learn from here, but also a lot of experts who can piece things apart.
I feed my completed posts into multiple LLMs for multiple rounds of reviews asking "Comment on this post as a member of $orangesite and lobste.rs". It frontloads critiques and helps build confidence for what I post, but that particular gauntlet also kills a lot of posts.
in the nicest way, i dont understand this
i say this because i can also emphasize with tropin - nowadays, it is understandable if not acceptable, as to why he would make such accusations without proof (also noting that he did not really mind if it were true -- "Despite the way that post is written (with LLMs or not)")
why delete the post? the accusation remains, the post doesnt provide any kind of proof of the negative, and says so itself. the author couldve written about how they felt about this situation without deleting the post. if this is to avoid harassment, i dont see how making another public post on the matter would help in any way
to be clear: i'm not upset about being critiqued. i actually welcome that,
posting on the internet is pretty much always going to result in unfair critiques. if you can not handle unfair critiques, maybe, for your own sake, you should avoid writing publicly overall
if you can not handle unfair critiques, maybe, for your own sake, you should avoid writing publicly overall
I'm sorry but this is basically victim blaming. "if you can't handle other people saying mean things, then your viewpoint isn't worth hearing from". Why are we silencing the people who did nothing wrong, instead of telling the assholes to knock it off??
I know we can't control the internet writ large, but at least on lobsters maybe we can stop throwing around unfounded accusations on every single post???
sorry, couldve phrased it better
"then your viewpoint isnt worth hearing from" is not what i intended to say at all -- the viewpoint may very well be very worth hearing from, but it may be in the interest of the mental health of the author themselves to refrain from posting. it is, of course, very unfortunate that this is a problem overall, but this situation specifically reads to me like an overreaction which is why i am saying this
This reads so much like “look at what they wore in that part of town”-style victim blaming that I don’t know where to start.
It’s amazing to think that we are now expected to self censor our own thoughts out of well-deserved fear that sharing those thoughts will be interpreted as ecosystem destroying plagiarism. Franz Kafka eat your heart out.
victim blaming requires blaming
i am not blaming the author, even if i do not understand their decisions, and think it is very unfortunate that they were unjustly called fake -- i wish them all the best and indeed would prefer if they didnt stop posting
i am only trying to offer advice because this seems like a very difficult situation for the author where it normally isnt for other people
i also already said all of this in the comment that you're replying to, so please actually engage with what im saying if you are going to reply; critique the advice instead of putting words in my mouth, and abusing therapy speak + overly dramatic metaphors. replying in this manner to a comment under a post about how someone being unjustly rude made someone feel bad is incredibly ironic
You are right to point this out. I clearly didn’t read your position clearly. Have a genuine apology on me, if you like. Thanks for taking the time to correct me.
One thing I've spent a lot of time thinking about with respect to my own blog/writing is: why do I do this? Am I writing just for myself? Am I writing for an audience? Who is in that audience? Who am I hoping is going to read this and what am I hoping that they will get out of it? And why do I care about that - what do I get out of it?
To be candid, for me a part of it is about attention, recognition, and ego. I feel like I have things to say, and being listened to does something for me. Part of it is an enjoyment of the writing process - an aesthetic desire to precisely pin down a feeling in language.
I do experience dissonance about it. When things I've written have gained attention, a lot of the comments I get are not really what I'm looking for. I think engaging with someone's writing sincerely and charitably is very hard to do even when you intend to, and most online commenters are lacking even the intention. And yet I still cross-post to lobsters and hn, and watch the read counter.
And I guess, being an absurdist, I accept the incoherence. The desire I feel is an evolutionary drive for recognition, reciprocity and relation in a real, physical community that's misfiring onto this abstract, sort-of-community-shaped thing that is blogging on the internet.
In any case, tying it back to the discussion at hand - this is the sort of thing I think about when I see people trying to tease apart whether coopi's is being reasonable. That if you post your writing on the internet then you should be prepared for certain reactions. That there is some homogeneity that we can expect of people - why they write, why they share it publicly, and how they ought to react. There is little about writing for an audience of a billion anonymous strangers that is reasonable.
I guess the only thing that help against such accusations is trust. Of course, you can't know or trust anybody on the web but human.json might help a bit.
This is such a shame. As someone who has used Nix a fair bit and dug into Guix more recently I quite liked the original post. When reading it, my "ai sensors" went off due to the em-dashes, but it seemed like the the text had a good amount of personality showing through that it was more likely just written by someone who appreciates and makes use of the whole grammatical toolkit.
I like reading things people write (like coopi's original article) and wish we didn't have to constantly second-guess whether an LLM spit out the text.