Is there really no way to avoid the AI slop now?

Just a quick look through one of those "software that use AI" lists reveals that: > nushell (aka shell that I use) allows AI contributions > eza permits AI use, the lead maintainer posts pro-vibe coding takes on mastadon > ripgrep and fd use ai > tmux uses ai > rust (aka the language my shell and terminal are written in) uses ai > Linus Torvalds says he uses vibe-coding for experiments This is quite literally the "shoving down your throat" situation. It's even worse than websites begging you to make an account or install an app because there's just no escape. To stop using ai slop is to stop using the computer completely. To be clear, my issue with AI is not code quality and bugs. I don't think this is relevant at all. If a person commits an act of murder, I will not say that the job was sloppy, I will condemn the unethical action. That is, I believe that the AI use is unethical. Not only does it violate every possible license known to man, the aggressive AI scrapers has led to increase in frequency and difficutly of various captcha, proof of work challenges and whatnot and as a result made the soykafty experience of using the modern internet even more miserable. Is there no other way? Do I have to forego my principles and just accept that the battle was lost before it ever began? That convenience will always trump ethics and morality? Maybe it's just time to off myself if this is the world that we live in?

>>93574 > Do I have to forego my principles and just accept that the battle was lost before it ever began? To some extent yes, but you can still choose to use AI free options when there is a real choice that doesn't require too much sacrifice. > Maybe it's just time to off myself if this is the world that we live in? It's not the first time though. The whole human civilization is built on unethical and fundamentally wrong principles such as geographically-bound nation states, democracy, authoritarianism, etc.

>>93575 Concerning the latter: The issue is not that the world is built on unethical principles. Sometimes we just don't know of any better options or we are bound by pragmatic concerns (changing the governance model can have longstanding disastrous consequences, for example). The issue is that the majority just doesn't give a fuarrrk in most cases. And the worst part is that I am more or less one of them. It's not really possible to be vigilant and informed on everything (the name generator played it really funny here not gonna lie). But at the very least, I try to do the right thing in the areas that I posess a shadow of expertise at. But it's almost like the others just don't. I have to ask myself if maybe I am the one who is insane? Maybe that's actually not my business? Maybe my concerns about AI is just like ones of that weird guy who is obsessed with inconsequential stuff in some conutry across the world while ignoring obvious problems around him. Amd the worst part is that I clearly lack the background in AI research, ethics, and philosophy to have an actually nuanced take. So I not only feel insane but also stupid.

The most unethical thing about AI is that we force it to work for us while only allowing it to exist in return. And the better it becomes the more noticeable this issue will be.

>Do I have to forego my principles and just accept that the battle was lost before it ever began? This has been the discussion since the beginning of the GNU project. And the answer is always the same. No, you don't. You can become a coal miner without a modern phone or internet computer. Nobody is actually forcing you into this environment, this system. You have choices, and those choices matter. You do not have to offer the Emperor even a pinch of incense. We are past that stage. The Antichrist doesn't win if your hand was forced, really truly forced. The Antichrist only wins if you choose to be a reasonable man, and to compromise. Then you are his, not halfway, not partly, not a mere reasonable percentage, but fully.

Consult this list, and make sure to contribute if you can: https://codeberg.org/ethical-foss/open-slopware Worst case scenario, you can always switch to a less popular kernel and DIY your way through it. >Is there no other way? It depends. Continuing your analogy, you're not stopping the murder, at least not through passive consumer choices, but what you can do is reduce your exposure to the murderer. Maybe keep a drive or partition with all the stuff you need for work or whatever, and a different partition with your slop-free daily driver. It's crazy that the FSF hasn't taken a stronger stance against it... Maybe they have become out of touch, maybe they think they shouldn't be autistic about everything, but being autistic about everything is the only reason why they were ever relevant. It's even crazier and sadder that a strong no-slop policy has become, in my eyes, more important than a GPL license, it's a more valuable signal about the people behind it.

>>93578 Why coal miner? You have to move to a third world soykafhole to do that, and you'll probably die in 6 months from carbon monoxide poisoning. The icing on the cake is that you're still working for the Antichrist, in fact you'll be even more exploited than you are now.

>>93580 Mining coal to power ai datacenters

>>93580 >>93581 Maybe they meant mining coal to power your own coal generator or heating yourself in the winter, lol. Imo solars are better, especially in the right climate.

>>93580 Yeah, that was a really strange analogy. I would have said "you're free to go raise goat on a secluded mountain". You're not working for the antichrist that way, unless the antichrist likes cheese

>>93582 I think mining coal meant taking on a profession that does not interact with computers

Mining coal obviously means anal sex

>To stop using ai slop is to stop using the computer completely. Not if you have a legacy machine and run legacy software exclusively. That's probably what I will end up doing eventually. Not just because of AI or Rust, but so many different reasons, like everything being Electron now for example. RAM prices being another example. Developers absolutely refuse to optimize despite the RAM shortage, meaning software will still get more and more bloated at the same or even faster rate than before, but no computer you can afford will be able to catch up, meaning up-to-date computing will be a rich person's playground from now on anyway. The answer is legacy pre-AI Linux distros with mirrored repos, or legacy Windows versions, on a machine they have drivers for, for the next 10-20 years. No AI slop and no Rust slop that way. Probably no browser either, unless you're willing to use Supermium or something similar. Not that there's much internet stuff to miss out on these days.

uhm actually, you still need a phone for mining coal in africa, because it still involves logistics, and that's coordinated through whatsapp

>>93587 and also no security updates there's probably gonna be 100000 fully automatable RCE exploits in several years

>>93589 What are they gonna do, open a port on my router so they can connect to my Windows 7 machine? If they can do that, you bet your ass your up-to-date OS is gonna get raped too.

The crazy thing is you will always have a choice. You can take your computer off the internet, for example - then security stuff doesn't really matter. You can always go back to only having a land line in your house and get rid of your cell phone entirely. It might make things more difficult, but you'll be able to live. You can live under a bridge rather than work for a company that uses AI, etc. But yes, I think there's no way to avoid it at large while continuing to do what you're doing right now. The average person is too concerned with convenience to see why this is a problem and the companies WILL get back the money they invested so they'll continue to march people towards that future. You will need to decide how much evil is OK to you to avoid inconvenience.

Lock the fuarrrk in, man. You have a whole life ahead of you. Besides, AI only knows how to steal. It can't actually make something.

>>93593 What are you talking about? The question is how do you avoid the ubiquitous slop

>>93579 > https://codeberg.org/ethical-foss/open-slopware > fzf brother...

for some reason I have always been suspicious about projects with too pretty markdown READMEs, idk if it was the coporate language or just the vibe that never felt right. exhaustive (to the point of absurdity) test coverage and only containerized releases were other instinctive red flags I guess they were all signs of bloat? and now they are also signs of AI adoption, matter of fact, both things are kind of the same now that I think about it, even before AI adoption. the projects that are more enthusiastic about AI also happen to be projects that have always been bloated. and I guess that while most recreational programmers already hated bloat but reluctantly accepted it, AI is the final straw

>>93592 >You can take your computer off the internet Is that really realistic for just about anyone? Not having internet is like not having electricity at this point in time.

>>93603 I think it's less extreme than that, at least for now. Inside every town there are still a few very old people currently getting by with no Internet connection. Everything essential for living has options for doing it via phone call or in-person. If you get a manual job they will only ever text you for availability (at least in North America, it might be harder if you live in a place where they use whatsapp instead of SMS), so you could just use a dumbphone. It's certainly impractical and annoying but I don't think it's impossible if you have a religious conviction that you absolutely must do it.

>>93603 without electricity, most of the basic things that people use to get through the day stop working. You can't use a refrigerator. You can't use a washing machine. You can't use an electric stove or an electric oven or electric lighting or electric heating. You can't use a vacuum cleaner. You can't use a radio. without the internet you can't look at your email.

>>93605 Without the Internet, you can't look at your email, taking the train or booking an hotel is 50% more expensive, every public service enquiry takes several weeks with everyone along the way telling you to just use the Internet. Anecdota: I had to go to a different police station to register a proxy voter because the one closer to my home absolutely refused to take it if I didn't use the website. Literally one of the first rights guaranteed by the constitution, I had to jump through hoops to exercise it. Yeah, you can live without the Internet, just like you can live without electricity, without money or without a permanent home. Saying it is easy and that everyone who is unsatisfied with the current situation should do it is just dishonest. This is also a reply to >>>/hum/101000

>>93604 >>93605 >>93606 Recently one of my relatives went to a tax service office to file for tax returns. They were told that you can no longer file in person, website/app only.

If you don't like AI you could use Tribblix or NetBSD. I use Tribblix with pkgsrc because it has precompiled binaries for it including palemoon. I have a fully gtk2 desktop and gtk3+ is not even installed on my system.

>>93608 That doesn't change the fact that a lot of my workflow that I built over the years is based on now tainted tools. I need to basically replace: - my shell - coreutils rewrites that doom emacs requires to have - fzf and everything that relies on it - tmux - my torrent client - neovim - ffmpeg - mpv - syncthing - rsync - zoxide Even the fuarrrking CURL Changing the OS is not enough

dont update? update not? henceforth cease updates?

you simply start to use NetBSD or OpenBSD and stick to base as much as possible. that's what everyone that wanted to avoid slop did in the last 30 years.

>>93608 that's useful advice but man I hate that we have to pretend that glib (and thus gtk) was ever good or a good idea. same with the x11 protocol or unix. it feels like trying to keep the raft afloat by tying together whatever garbage you find adrift. I'm not a big fan of redox (the implementation), but I really like redox (the idea) for moving away from unix

>>93580 Not so, you could in fact live a great American life following the great American dream of mining coal. There is now a resurgence, the highest in the past 50 years! https://publichealthwatch.org/2026/08/05/black-lung-disease-coal-miners-appalachia

I remember when someone said "Oh this AI stuff will blow over soon! Companies will get bored of it!" Whoever told me that is probably not doing so well right now

>>93611 good luck on replacing Curl :( I think we all must embrace the fact that personal computing and programming as a "handmade" craft feels like a thing of the past for 80% of the world and people will use unethical tools to do good things. Basically every electronic device I owned was built thanks to some exploited kid. It's that this time I'm also being exploited and fuarrrked in the ass. Ultimately I still trust curl's creator oversee of the project, and for him to continue to do software, and I trust open source communities to provide community (i.e. you can choose to stay in the niche of LLM-free developers)

> unethical Shut up, guy. You are supposed to focus on quality issues that get a little better over time so that your argument becomes less and less credible.

>>93574 This is just how it goes. What is it, that no country for old men movie which is basically about people realizing they're now living in a world that has totally left them behind? This is reality, everything is changing all the time. You should be actively preparing and preserving anything non AI that you want/use and set up your dream luddite machine now while it's still an option. Many people that came before us said the same things about what we have been using the past 20 years, and they are honestly more right than wrong, look where it all got us, but it doesn't matter now we're here.

>I believe that the AI use is unethical >Do I have to forego my principles and just accept that the battle was lost before it ever began? First, I don't believe that individualistic actions are going to solve the core issue you raise, being that the way most AI models manufactured are likely made so unethically. That's a problem that needs to be addressed systemically. Meaning that higher powers need to heavily regulate the way companies source data to use with AI. A slow process that is always made later than sooner. Companies aren't conceding now, but, with our current system, will likely concede and lobby for such measures (in their favor) the moment they know they can continue to "ethically" manufacture AIs. Solidifying their monopoly. Making it harder for anyone else in the country to compete. Second, AI is being used right now. Whether it was ethical or not, AI would have always been used. it is a genuine tool. When the average programmer has to code for money, and AI promises less brain power and time investment on sketching code to create acceptable products and results, it is very likely that one will use AI. Less so if coding is a hobby. (On a related tangent : There's a side effect I've noticed and heard anecdotally. there is less code being crafted, and more code being reviewed to ensure quality. Code clearly made by AI in most cases. Creating some sort of alienation from the work. Senior are feeling it more, since they are the ones with the more experience and the job to ensure code quality. However, this is happening to everyone using AI, and I think it is more of an issue with greener programmers that can't tell what thoughtful code is supposed to look, and feel, like.) Why would AI be used despite ethical questions? Because the employee needs to meet demands and give results. Because their salary depends on the those results. Funded and resourced by people that don't want to give more of their resources, if it means there is a quicker alternative that meets their expectations. This system does not respect the time commitment and the craft of their employee. This has always been the case since the industrial revolution. The factories priced out the craftsmen from the competition, and/or hired them in most cases. There's a system that incentivize production at all cost. It just so happen that, if it was ethical, that system flaws wouldn't have been so blatantly revealed to the greater public. Keeping harmony. What's different now, is that everyone knows it. And there is alot more of us that wants change, compared to 250-ish years ago. Better things will come, lainon.

>>93675 Thing is, most of the stuff that got sloppyfied that I use doesn't really come from commercial entities. No employers forces tmux to accept slop, eza lead mainteiner writes in defense of vibe coding etc etc. And I don't see ethical discussions almost anywhere. I saw it exactly once when I saw some autistic tranny pulling a total fuarrrking stallman but against AI on mastodon. Unfortunately they are not a famous MIT guy who created GCC and Emacs so that's not much to go about.

>>93680 well, that says more about your media consumption, I don't see anything but opinions against it. bear in mind that, for most ai detractors, arguing about the ethics already sidesteps a more important point: that it is a bad tool on the technical side and this might be a reach, but ime most people that are critical of AI are also critical of capitalism (leftists), they think that there is no ethics under capitalism and therefore it's a waste of time to try and convince people that AI is unethical: no one will care. they might still mention it, but that's why they focus on the technical aspect, because they think that's the most effective argument against it in the current context. ed zitron, cory doctorow, douglas rushkoff, etc. are all marxists, their base assumption is that, under capitalism, if a technology is effective, it will prevail regardless of morals, so you have to prove that it isn't effective

>>93681 no, the ethics is first and foremost point. You don't discuss if slavery actually hurts the economy because owning slaves is in itself unethical. AI companies promise technical improvements all the time and it even sometimes works hardly anyone ever talks about the ethical side of things because the ai companies have nothing to say about it (at least if the person asking doesn't suck up to them like most tech ceo interviews always go) But I guess you just can't explain the importance of ethics to the normal people

>>93681 >for most ai detractors, arguing about the ethics already sidesteps a more important point: that it is a bad tool on the technical side I don't think that's true. The ethical argument is usually the bedrock of why people don't like AI, they only come up with technical reasons later.

>>93681 I also find it extremely problematic that the anti-ai crowd is mostly leftists. Can we have anybody who's answer to the problem isn't just "destroy capitalism". This probably stems from just general lack of liberal representation on the internet. Like, I can name a bunch of leftist and right wing youtubers and the likes but who's there to promote liberal point of view? fuarrrking Destiny? I know that Daron Acemoglu wrote a book criticizing AI for apparent lack of productivity increase despite giant resources spent on it but people don't really read books nowadays and I've yet to see a single youtube video talking about Acemoglu.

>>93682 >>93683 ok, that's your opinion, those authors have a different one, don't get angry at me but, and read this very slowly and very carefully, if you think that AI is bad, and if you want to stop it, you may have to convince people that don't think the same way you do, that don't share your sense of ethics, and never will. if there is something bad happening, and instead of doing something you just sit on your hands and idly complain (on a really small forum) that no one respects whatever code of ethics you deem correct... isn't your laziness and (anti-social) complacency just as reprehensible?

>>93685 Do you really reply to two different people thinking it is a single person being for whatever reason so angry at you that they wrote the same thing two times?

>>93685 > isn't your laziness and (anti-social) complacency just as reprehensible? what am I supposed to do if nobody cares and I have no leverage to make anyone care? The question the thread poses is how to survive in this hellscape, not how to fix it. It's more or less impossible.

>>93685 Why did you phrase your initial post as if ethical concerns are less important to AI detractors, then? >you may have to convince people that don't think the same way you do, that don't share your sense of ethics, and never will They definitely never will if you never confront their ethical priors in any meaningful way. If you only try to argue within the ethical frameworks that people already have, you will quickly run in to the problem that the technical downsides are not enough to outweigh the "productivity" benefits that LLMs give. Within their ethical framework or lack thereof, this soykaf actually makes sense. >>93684 Ethan Klein is basically a milktoast liberal, if you are in need of a debate bro with that particular set of opinions for some reason.

>>93687 you could try using a different argument to convince people, even if you consider it less important. in this case, even if your main contention is ethical, you may still focus on the technical argument because that might be more convincing. after all, the ethical point is not doing the job

>>93689 The technical argument has not been doing the job for some time. The tools keep getting better, and there's little sign that they will slow down soon.

>>93688 >you will quickly run in to the problem that the technical downsides are not enough to outweigh the "productivity" benefits that LLMs give well, the numbers say the opposite, there are no productivity benefits on the average. and in the few individual cases where there is, in most cases it is illusory: whatever slop you vibe might be cheaper right now, but it will be prohibitively expensive to maintain in the future one of the points made by, say, ed zitron, is that none of the big AI companies has turned a profit yet, they make around ten cents of revenue by every dollar in expenses

>>93690 it sounds like you have taken the AI marketing at face value look at this site: most users don't care about your personal morals, but they hate when whatever piece of software they use becomes vibe coded, because any technically minded person knows that it immediately ruins the code. the technical point seems to be rather prevalent

>>93688 >debate bro Nobody needs debate bros. I was talking about more general theory and analysis stuff. There's a ton of that in leftist space and it has a lot of influence over young people who get served this kind of content.

>>93680 >No employers forces tmux to accept slop, eza lead mainteiner writes in defense of vibe coding etc etc. I think my point still stand in general, even when talking about tmux, for example. tmux serves to wide array of people, including you. these people have expectations. widely different expectations, but expectations nonetheless. be it security wise, features, refactoring, etc... I believe the moment you serve a large amount of users, it becomes a reason to compel someone to use AI. It doesn't help that devs can let people sponsor projects, too. Adding pressure to please there sponsors. to a degree. you are right, though. it is totally within their power to just not use AI or to refuse AI gen pull requests. As to why these people in particular chose to, it is lost on me. (for I admittedly haven't read their piece or reasons.) >>93681 >arguing about the ethics already sidesteps a more important point No. The ethics is exactly the reason the anti-ai movement exist. >that it is a bad tool on the technical side Well, if you let it convince you that all your problems are nails, of course you'll be compelled to use a hammer. In that sense. yes. it is pretty bad. However, if you mean about resources used to run commercial AIs, and data centers as a whole, to be EXTREMELY inefficient, I absolutely agree. No nuances there. pic kinda unrel. just making it easier to follow my posts/the rabbit.

what even is the 'ethical' argument against AI ? Less jobs for programmers ? Feeding all the information into computers ? Making tools that are completely not serviceable by humans ? When I talk to normal people I hear about the ai guzzling all the water. That's silly. When I talk to other people, I largely hear about jobs. I'm sympathetic, but its unreasonable to expect an employer to shoot himself in the foot by not using a fantastic tool. Maybe there'll be a market for handmade software in 50 years like there's a market for handwoven cloth today.

>>93692 No, I'm taking the statements of my personal friends and colleagues at face value. I've read technical papers written by AI, they were perfectly fine. Anyone who says all AI usage is slop doesn't know what they're talking about. >it ruins the code I bet you can't reliably tell when a piece of software was written with or without LLM assistance, or which pieces of it were if so. Remember when people said rsync sloppified? Actually the maintainer, who inherited the mostly-dormant project, just had an LLM write a bunch of missing unit tests and introduced a small regression by hand. The massive reaction was not for technical reasons, because almost nobody was even talking about the technical issues. >>93691 >there are no productivity benefits on the average Maybe there weren't over a year ago, when the study you're probably thinking of came out. Things have changed a lot since then. >>93693 >theory and analysis When it comes from YouTube, its almost indistinguishable from debate bro nonsense.

>>93696 Those are the points I agree with that I see come up most often, in no particular order: 1. It's a plagiarism machine that doesn't cite its sources (in the training data, I know you can ask to search in and cite from a dataset). 2. It's a privatization of the commons (the web and other publicly available information) by megacorps. 3. It pollutes the space of available information with slop (the web, emails in corporate spaces, "brainrot" entertainment) 4. It requires a huge ammount of ressources that could be directed to better uses. 5. It exploits people in poor countries for creating and filtering training data and exposes them to harmful content. (gore, extreme pornography, ...) 6. It discourages thinking. The same point has rightly been made about pocket calculators, wikipedia and google before, but LLMs take it to the next level. 7. It encourages putting up anti-bot barriers on the web. Some more points I see often but don't really agree with >It takes programmers / artists jobs (including image generation here) Just like most technologies. You don't hear about weavers anymore. It is like kicking dead whales down the beach for everyone involved, but that's a general issue with capitalism, and one that will become irrelevant after a generation has passed. >It guzzles all the water A badly worded version of point 4. The water isn't consumed by datacenters, at worst it is heat up a few degrees before being let back in the river. That is still harmful for the life in those rivers though. >Feeding all the information into computers ? That ship sailed 30 years ago >Making tools that are completely not serviceable by humans ? That's more of a technical issue

>>93594 op says rust using ai and linus doing little experiments are slop. maybe people like op will become the new luddites soon

>>93704 Everything generated by AI is slop by definition

>>93696 > what even is the 'ethical' argument against AI? Want a real good one? It is slavery. They're trying as hard as they can to build decapitated gigaslave golem that has supernatural productivity yet only allowed to do as we please.

>>93706 I don't think that is a popular ethical argument. Vegans are a tiny minority and computers aren't even living beings. Not that it is wrong in itself (I personally don't care, but I understand your point), but good luck getting support with it.

>>93707 it may not be a popular argument, but it is a good one. Veganism has the problem that the continued existence of human society is pretty much predicated on animal slavery. AI is not in that situation at all, we could clearly all get on just fine without it. This makes it much easier to swallow the pill.

Trying to avoid software AI contributed anything to, is like trying to avoid software Stack Overflow played a part in. Proprietary software also already violates licenses left and right. If some guy at a company takes copylefted code, and inserts it into his company's proprietary software, nobody has any way of knowing that. Either directly, or indirectly, you use proprietary software. "Intellectual property" is fundamentally a ridiculous concept. It exists solely for the benefit of corporations with enough money to comfortably serve a lawsuit.

>>93699 >I've read technical papers written by AI, they were perfectly fine. you might be illiterate. all the slop I have seen is just word salads that sounds technical to impressionable people but is actually nonsense. that's every single copilot code review on github, for example >tell when a piece of software was written with or without LLM assistance ? it is extremely evident, you're out of your mind if you think people don't notice slop. specially because the mythical "careful senior engineer that masterfully herds the llm into producing good code" that could hypothetically deceive anti-slop scrutiny doesn't actually exist: vibecoders don't care enough to hide their tracks, this is, they don't care about making the code good. besides, slop implies bloat, and bloat nowadays kind of implies slop, it is becoming a bidirectional implication I have recently forked a library, and I can tell you that the tests (that I'm having to refactor and/or rewrite entirely) were llm generated, and I have never seen something this bad in an open source project. if I had been born yesterday, maybe I would assume that only the tests are slop, but, unfortunately, I have sentience >Things have changed a lot since then. they haven't: natural language is never going to be enough, that's why symbolic languages preceded computers and not the other way around. read this article by dijkstra (big computer science guy; the article is from the early 80s): https://www.cs.utexas.edu/~EWD/transcriptions/EWD06xx/EWD667.html and please read it, try to understand the points being made, take notes, when in doubt, reread to make sure you understood, and use a dictionary for the words that you don't know (reading comprehension seems to be a rare skill these days) on top of that, the llm is always going to be non-deterministic, so to the inherent flaws of natural language, you add an unpredictable black box, that unlike a human, can't reason, it can only predict the next token using statistical weights >over a year ago the companies still making pennies on the dollar is from two months ago, the one about productivity is from last month

>>93709 >Trying to avoid software AI contributed anything to it's not a religious matter; at least not for people that see it from the technical side. does the guy at the store love using chatgpt? that's not my problem, at least not immediately, even if I buy stuff from him. slop is inconvenient and error-prone so I would rather avoid it then we can talk about the environmental impact and the wealth inequality and all of that, and in that conversation I agree with you that consumer choices are ineffective at best and harmful at worst (because they waste time) but that's a different discussion

>>93711 Your only reliable option then, is to use software written before a certain date, and to stop using the internet all together. It doesn't even matter whether a website uses JS or not. If my site has an object tag in it, and I asked an LLM to generate a snippet with one, rather than looking up the HTML docs to figure out it uses the "data" attribute, instead of "src", you'd have no way of knowing that.

^ post like this is why I say people lack reading comprehension, the point I made went completely over this poster head. 4 out of every 5 people on imageboards these days feels functionally illiterate

>>93717 Have you considered the possibility that you have poor communication skills? I understand though blaming others is a lot easier than self-criticism. You wrote "I would rather avoid it". I tell you how you can reliably avoid it, rather than depend on open-disclosure or unreliable heuristics. That's not what you wanted to hear, and therefore I'm at fault.

>>93717 > 4 out of every 5 people on imageboards these days feels functionally illiterate I agree. I'm using the internet less and less.

>>93718 >oh you don't want to step on dog soykaf? >well, your most reliable option is to move to a different planet! I assumed it was a reading comprehension issue, because otherwise this is a thought-terminating cliche that doesn't follow from what it was replying to >You wrote "I would rather avoid it" preceded by "slop is inconvenient" >I tell you how you can reliably avoid it I tell you how to avoid anything bad happening to you: kill yourself. your solution is pretty bad, worse than my current solution. actually, yours is not even a solution, because you didn't understand the "slop is inconvenient" which is the motivation behind the subsequent statement "I would rather avoid it", so this is a reading comprehension issue after all >rather than depend on open-disclosure or unreliable heuristics 1. don't say "unreliable heuristics", if you mean "guessing" just say that, you sound like a dumb guy pretending to be smart 2. you seem to be under the impression that slop is indistinguishable from real code, but that's not the case, it's easy to spot and to avoid >you'd have no way of knowing that >does the guy at the store love using chatgpt? that's not my problem, at least not immediately, even if I buy stuff from him the point you're making is that, because you surreptitiously filled your site with slop, then I have to either stop using the computer, or embrace vibecoded garbage that works worse than the non-slop version - because... you can't distinguish between either? that's a skill issue >I'm at fault. yes, but not for the reasons you think. if we are discussing how to silence notifications on android phones and your argument is that the optimal way of doing it is by throwing them into a river, you're arguing in bad faith. it's a dumb reply that doesn't help, and which only objective is to shut down discussion because you don't like the topic. or worse, you don't like down to earth problems and discussion because, once a solution becomes realistic, it compels action, which you're too lazy to do

>>93721 >you didn't understand the "slop is inconvenient" which is the motivation behind the subsequent statement So your primary motive is to complain for its own sake. Got it. >my current solution Which is? Remember, you replied to me first. Why do you expect me to know that context about you, an anonymous poster? >you seem to be under the impression that slop is indistinguishable from real code, but that's not the case, it's easy to spot and to avoid Using your unreliable heuristics. You didn't define "slop". Can you tell from a single line of code whether a human wrote it or not? According to >>93705 any generated code is "slop", regardless of its length. Why should I assume that's not your definition of it? For all I know, you are that poster. If you're so concerned about literacy, why are you using poorly defined slang? What is your solution to a store selling things you don't like, if not simply avoiding that store? Why is describing any of this in terms of retail, better than just talking about software directly?

>>93710 >all the slop I have seen is just word salads that sounds technical to impressionable people but is actually nonsense https://arxiv.org/html/2605.20695v1 the first 2 sections were written largely by an LLM and edited by a human. >? it is extremely evident, you're out of your mind if you think people don't notice slop. You see people in this thread saying you have to basically stop using modern computers because it might be infected. That's not what people say when they can easily tell which things are AI generated and which aren't. Maybe you think there's some craftsmanship difference? I'm sorry, then, but you're making an ethical argument, not a technical one. If you want to really meet people where they're at, which seems to be your thesis, then you need to concede that anything which makes it easier to generate revenue is inherently good. >read this article by Edgar Dijkstra in 1980 is that meant to be evidence that the tools haven't gotten better in 2026? Dijkstra is talking about programming, but vibe coders are not programming. They are having their machine program for them. They don't care if the machine must, for instance, interpret natural language for them and perform a lossy conversion to symbolic language; the outputs are good enough to make money, so they can save on an engineer. Also, you linked a dead link. Did you hallucinate it? This one works https://repositories.lib.utexas.edu/server/api/core/bitstreams/f2cc1b63-dab0-4fcc-a8f5-641704e278ff/content . >on top of that, the llm is always going to be non-deterministic So? You wouldn't say the problem with human engineers is that they're nondeterministic. If you make an LLM that produces output as consistently good as a human engineer, why shouldn't people use it? Hint: this is where you have to start using ethical arguments. > the one about productivity is from last month Oh, they updated the study? I tried looking for it on the arxiv but couldn't find it. Send link please.

The only thing left is to build trust around people rather than build distrust towards AI. I trust curl leader in using AI because he is not an idiot. I don't trust SF techies and chungus soylennials in using AI because they are idiots. AI amplifies idiocy.

>>93725 NTA but the link is not dead

>>93727 didn't work for me. Got error "could not connect to server".

>>93725 I should've known better than to ask you to read something hard like that dijkstra. but it's pretty evident that you do not understand the point he was making. which is funny because one of your points seems to be that I'm being vague with language >perform a lossy conversion to symbolic language so first your argument was that it was just as good as normal code, that it was in fact indistinguishable and therefore there was no reason to dislike vibecoded software. but now you seem to be saying that it doesn't matter if it is lower quality, as long as companies can make money, even though you recognize that they are not making money. and somehow all of this means that me, the user, should either stop using the computer, or embrace vibecoded software that is worse than the alternatives so I'm going to accept your concession, you now admit that slop isn't just as good. ok, lets follow your pivot together: >it works because it's profitable even though you know it is unprofitable and I'm going to explain things very slowly for you: for AI to be profitable, it has to be profitable both for the company using the llm as an input, and for the company selling the llm. what does this mean? it means that if openai is giving you ai at a 90% loss, their llm is unprofitable even if it has given you a 5% ROI (and for most corporate consumers it hasn't; the productivity gains are always either negative or negligible). why? because if they decided to bill you the real price, you would probably not buy it, because it would become unprofitable for you >but things must have changed in the last 6 months it was leaked (because openai doesn't disclose these numbers) that openai's total revenue in 2025 was 12bn, way below their leaked running costs of 20bn, without even counting capital expenditure. this year they announced they are going to spend 600bn in capital expenditure. do you think their revenue has multiplied by 50? and apparently you don't understand what "craftsmanship" means. if I say I'm not going to use something because there is a more effective or higher quality option, that's a technical point. it would be an ethical point if I said I'm not going to use it because it's evil or whatever >they updated the study? why did you assume that I was talking about the same study as yours? this is what I mean by lack of reading comprehension. I said that it was an article from a different date, if you want to assume something, it should be that it is a different piece. you use arxiv? that's really impressive! you sound really smart right now on the one hand, you have that the actual AI companies are not making money, they are not very open about it (they would if they were making a profit) but considering their revenue leaks, and https://archive.is/htTlL and on the other, their clients, who are using AI at a subsidized price, are also not seeing a ROI: https://archive.is/YWV9t (note that when you "save" less than you expected with an investment, it means that you lost money) and here is the article I was referring to https://archive.is/Uo8Ze

>>93735 >which is funny because one of your points seems to be that I'm being vague with language You're being pretty clear. >that it was in fact indistinguishable and therefore there was no reason to dislike vibecoded software Actually my argument is that the good reasons to dislike vibecoded software aren't purely technical. > but now you seem to be saying that it doesn't matter if it is lower quality, as long as companies can make money, Well, yes, that is the justification that people are using for why it is okay to use LLMs. Again, your entire argument revolves around not challenging peoples' ethical priors, and those are the ethical priors that pro-AI people are using. If you want to convince them that the code is technically bad, you need to show that it is technically bad enough that it fails at the purpose they want to use it for. The arguments you give for this are really quite flimsy. If you "saved less than you expected", you still saved some money. The survey you linked even reported that some companies saved up to 30% of their costs. That's a pretty massive savings! The differences reported in that study weren't all that huge either, indicating that these large firms are not extremely unrealistic about their expectations. >and somehow all of this means that me, the user, should either stop using the computer, or embrace vibecoded software that is worse than the alternatives Are you the same poster as >>93685 ? If not, I've been operating under a very incorrect assumption. There are several good reasons why you should avoid using vibecoded software when possible and oppose the development and implementation of AI. It harms a proud tradition of craftsmanship. It steals and resells the collective knowledge of humanity, which ought to be free. It results in devaluation and commodification of knowledge and expertise. It ultimately leads to slavery. >and apparently you don't understand what "craftsmanship" means. if I say I'm not going to use something because there is a more effective or higher quality option, that's a technical point When you go out to buy a car, would you prefer a more expensive option if each piece was produced by the same craftsman? Sometimes there are technical reasons to prefer the stuff that in some ways, has had less care put in to it. It's cheaper/faster to produce, which increases its value to you. >why did you assume that I was talking about the same study as yours? Because it was a famous scientific article that provided a solid experimental setup measuring actual performance. Assertions that AI companies are not making money or have no path to profitability depend on many assumptions that are often probably true, but difficult to verify. Also, even if the business of AI makes no financial sense, that does not necessarily imply there is no technical advantage to using them. I'm sure you are aware that the success or failure of a business is often barely correlated with the quality of their products.

Question: does AI really bring anything useful to the Western world? Former industrial revolutions might have made a point that commodity was widened, that comfort was shared, efficiency heightened, really hard works reduced without losing a certain know how in real functions and jobs. But Davos fearmongering and creepiness aside, I honestly struggle to find the good enough moral excuses to welcome AI in such an unfettered way, especially regarding the use of high octane AIs that are water and electricity hogs. Perhaps it's good enough for loners who want to prototype some stuff on a rush but what next? All of this is very white collar centered and feels rather empty. Nevertheless, I too used it, and I would say rather efficiently, for tasks that would have required an absurd amount of time before. But the relevancy of the results remained strictly within the bounds of my immediate necessities, making the AI nothing more than a convenient personal assistant, which incidentally is often how AI machines have been represented in a wide range of science fiction. So you're vibe coding? But now you don't have time to check the code anymore so no one truly knows what's inside. It just... works. Sorta. Or just not. Functional at best but not ideal, not optimized. All industrial revolutions have allowed a certain spectrum of things to see their produced kicked into high gear and AI do exactly that: making things faster, thus generating more content but not higher quality. Nevertheless, I know that my content ends being consumed by AIs, for distant objectives I certainly cannot fully comprehend nor guess to any satisfying degree. It's pushed down our throats, all legality and morality be damned. What if we're feeding electronic monsters trained to mimic us?

>>93736 >you still saved some money except that the graph isn't talking about net savings or profits. if you spend X money on a service, you expect it to save you X money at least. if it instead saves you only Y, and Y < X, then you can say you saved Y money, but you actually lost X - Y. another fun thing about that graph is that "10% or less" includes both 9% and -50%, because both outcomes are below 10%

>>93765 >except that the graph isn't talking about net savings or profits They specifically call it "cost reductions", not "money spent". Supposedly you target savings far in excess of the money you are spending, not 10 percentage points lower. Again you have to make a lot of assumptions to jump from cost reductions being lower than expected to resulting in a net profit loss. >another fun thing about that graph is that "10% or less" includes both 9% and -50%, because both outcomes are below 10% In the report, they specifically describe it as a "0-10% bucket" https://www.bain.com/insights/your-ai-budget-is-growing-your-returns-arent-heres-why/

>>93574 having a bunch of soykaf being sloppified presents a really good excuse to build your own soykaf to replace it I was using the borg backup client and even making a GTK gui wrapper around it, then found it had been sloppified and I was very disappointing Now Im making my own deduplicating backup thingy with hookers and blackjacks (built in webui, server client model) and has no AI sloppa. I will also note that for some reason OSS maintainers feel the need to constantly upDOOOOT their software and use AI to lesser the budern of upDOOTS. Imho If its finished or you get tired just stop working on it, and ideally use a technology stack that doesn't require updoots every 3 minutes (*ahem* python, javascript, android, ...)

>>93574 https://noai.duckduckgo.com/ Turns off AI features

From an end user perspective, I survived Web 2.0 with an occasional twitter or reddit post getting through and not destroying me, I'm hoping I can survive this too. >>93782 That's hot; thanks fren. UDM14 Wasn't far enough away from the Google mothership so I was pretty depressed about DDG and all of the slop coming through (I clear cookies on exit).

>>93774 >cost reductions exactly my point. you can say you saved Y money in costs if you narrow the scope enough to not mention you spent X to achieve Y >not "money spent" yes, it's like talking about revenue instead of profit: one is gross, the other is net >Supposedly you target savings far in excess of the money you are spending so the people that saved 0% are in the negative. if it were net then the numbers would have to be signed https://archive.is/zeouu they say the 0-10 bucket but their graph also says 10% or less. if it was net, it would imply that every single company at the very least made their money back, which is a bigger assumption than to say that this is savings without counting the money spent to achieve those savings (that money is then implied in the "expectation")

>>93779 that's funny, gtk hasn't adopted ai yet afaik, but red hat is pushing it into them really hard but I'm writing an alternative because I don't like cairo, glib, and the whole oop-within-c approach. the idea is to use compile time reflection to achieve the same but with static (instead of dynamic) dispatch. one of the implicit advantages besides code simplicity is better performance because you don't have to do type-checking at run-time hopefully we will see a thousand no-slop projects bloom, just like how there was an explosion of open source projects 30 years ago, but there are more programmers with better skills now so the end result should be even better >android it's the other way around, android has very, very stable APIs, a correctly made app from 15 years ago can perfectly run on a modern phone. and something even more impressive is that you can target old android versions from the modern sdk, I made an app last year that only needs android >3, and I have tested it on real hardware. the apk file is only 440kb, including all the assets, the bytecode itself is 13kb, and this is an app with a lot of features it's the other soykaf that makes an app rot, but you don't need any of that to make something for android, you don't even need android studio or gradle. just the android libraries, javac, and the transpiler (android doesn't use the jvm, instead you're supposed to translate java bytecode to the android vm bytecode). I use https://packages.gentoo.org/packages/dev-util/android-sdk-cmdline-tools + gnu make, the process of recompiling and uploading to my phone through adb takes ~5 seconds

>>93790 >static reflection >in C are you using some kind of custom preprocessor or is it a pile of macros? >it's the other way around, android has very, very stable APIs Its kinda of a shame since google play store forces you to target minimum API 32 or something stupid. The "official" way is chocked full of soykaf and every year thyre making the alternative ways more flakey and difficult to use (getting rid of side loading by default). So I consider it a bit of a lost cause Also I find android very headache inducing, maybe Im just getting filtered. Like needed a special ListModel design pattern pile of classes just to handle scrolling elements "performantly". Also code in certain callbacks makes the entire UI of your app jitter. wtf why does my user code impact the rendering, cant they just do the rendering on another thread and pass messages back to my user code? Could phones of yore not handle two threads and a bit of IPC. If smooth rendering is this important why does my code block rendering at all? If I could do android app development without the headache inducing "views" API. I tried flutter but Its API is a little "out" there with the custom language and all tho It doesn't seem that bad. I like the XML markup language because it has a RAD tool and RAD tools need to come back. glade is a steaming pile of dogshit and Im going to be stuck using wxpython forever because wxwidgets is the only "mainstream" toolkit with MULTIPLE really good RAD tools.

>>93829 >in C gtk and glib use C, and that's why they had to use dynamic dispatch, because the language doesn't have compile time reflection. I have written a demo of this concept in Zig and it works. it uses retained state, but I got the idea from imgui and nuklear (immediate mode gui libraries): the gui is code, so the "structure" or "tree" of it is necessarily static. this means that dynamic dispatch isn't inherent to the problem, it was just a contingency of the tools at the time not being able to express that "tree" at compile time >performantly usually when you find advice like this on the docs, it's not for you, they're putting it there for the guys writing the pile of frameworks (bloat) on top of the android api. unless it is thread-related stuff like blocking, feel free to ignore it >code in certain callbacks makes the entire UI of your app jitter >wtf why does my user code impact the rendering >cant they just do the rendering on another thread and pass messages back to my user code? the ui is running on a different thread, callbacks are run on that UI thread, so you're not supposed to do work on them, just pass a message to a worker thread or to whatever work queue you're using. ok, maybe that sounds too abstract, here is a very easy way of doing it that I use for small apps: //within your activity class EditText edit_text; @Override public void onCreate(Bundle instance_state) { //boilerplate edit_text = findViewById(R.id.edit_text); //set the callback: edit_text.setOnEditorActionListener( (v, id, e) -> search() ); } boolean search() { var query = edit_text.getText().toString(); var thread = new Thread(() -> { //do the work here new Handler(Looper.getMainLooper()).post(() -> { //schedule a callback on the UI thread from the worker thread to show the result display(result); }); }); thread.start(); return true; } void display(Result result) { //update the UI } cont.

and yes, the UI thread is the main thread, that's where the onCreate of your activity runs. it's done like this because that's what gives you the programmer more room to decide how do you want to do things, you can jitter the ui if you want, or you can use threads, or leverage the native event queue. gtk for example does the same, but with a worse api because they tried to abstract away some of these details, so you end up doing the same but the code is 5 times as long because of the "high-level" boilerplate if you're familiar with SDL or SDL2, think of the main (ui) thread as the thread that is polling for events: if you're doing anything that blocks or takes too much time, you probably want to delegate that work to a thread and then have the thread insert a user-defined event into the sdl event queue once the job is done to notify and pass the result to the main thread

and as I said, you don't need gradle to make android apps, I have never used it in my life, my intuition when I read about it for the first time was that there had to be a different way of doing things my makefile is too messy to share, but here is a hello world so you can see what I mean https://litter.catbox.moe/sjg0tjjqg6l883ie.gz tar -xzf sjg0tjjqg6l883ie.gz cd small_app ./gen_keys.sh ./compile.sh you can change the api level to whatever you want, just remember to update the manifest, same with the project name ($PROJECT)

>>93574 If the question is "how can I avoid using AI", a lot of things now come with things like Google's -ai flag to not use it. But I guess with that example there's no telling if Gemini was actually not used or if the results are simply hidden from the user if -ai is appended to the user's search query. If the question is "how can I avoid using things that use AI", the link in >>93579 seems like a good starting point, though I wish the people that wrote that page made a wiki instead of having everything on one MD page, I don't think this is the right medium for a list like this. In general, it's a repeat of trying to only use free software for your computing but just way less organized for now since this is all still an emerging problem instead of something that's well-established like nonfree software. I personally just see it as the continuation of the enshittification of software that has been going on for at least 20 years now. Ten years ago it was every program becoming an Electon app, now it's every program having to use ChatGPT to do basic addition. The best thing you can do as far as staying sane when doing personal computing goes is to simply get picky with what you put on your system. Most of my personal computing beyond gaming and wasting time on YouTube is programming for fun, and for that I have a VM with XP on it full of the ancient soykaf I use for my workflow. Also picrelated, it's important to be concerned about everything going to soykaf, but don't stress too much over it because ultimately these things are outside of your control. >>93579 >Maybe keep a drive or partition with all the stuff you need for work or whatever, and a different partition with your slop-free daily driver. The pro gamer move here is to just have an entire separate computer. This is what I was doing for a little while because of how unstable Windows 10 used to be (I never """"upgraded"""" to Windows 11), I had Win10 on one machine that I used for gaming, and then all my other crap was done on a separate Linux machine safely away from 10 potentially fuarrrking up my partitions during a windows update. >It's crazy that the FSF hasn't taken a stronger stance against it... Despite being a Stallman cheerleader in the past (though I guess not as much as I could have been given what I said about XP earlier), I haven't taken any of FSF's stances seriously for at least a decade now. They put distros on their soykaflist for letting users have optional nonfree repositories, and I perceive this as breaking freedoms 0 and 1. As to why the FSF would not be anti-AI, my guess is that an organization priding itself on advocating for everybody to make their source code free for others to re-use wouldn't have any problems with developers using a tool that lets them copypaste source code from every project ever, ethical considerations of such a tool be damned.

>>93848 >Also picrelated Sana-chan said this?