Why did it die?
easier to ask ChatGPT than some toxic neckbeard who will be like "why do you do that? You don't need that" / "that's bad practice / deprecated / obsolete / anti-pattern, etc"
all questions have been answered
>>108764925Because "technically correct" actually is not the best kind of correct in a low-trust environment.
>>108764938This is a good answer. SO was always a last resort. It's good for "how do I split a string in [language]", but you quickly grow out of that beginner area. Anything more complex and you're better off RTFM. Anything that's more SWE related and you should avoid SO and instead look for any of the other websites from Stack Exchange.
>>108764925there's no way it's that bad now
>>108764938I wish SO was still a thing. AI is terrible at answering questions that havent been answered. Half the time it hallucinates and half the time it gives you a related answer that doesnt solve the problem.
>>108765068>AI is terrible at answering questions that havent been answeredThis is very likely a skill issue, unless you are a scientist trying to get AI to solve an unsolved physics problem.
>>108765080NPC reply
>>108764937>chatbots got most of their code question answers from datamining SO>SO dies in favor of chatbots>Where do chatbots get their new answers from now?The next decade is going to be fun
>>108764937I also love how they pretend to have no idea what you're talking about because you didn't phrase your question in the most specific way as if they have absolutely no ability to use context from the rest of your question. Fucking obnoxious.
is the website actually dead? like did they shut it down?
>>108764945>how do I split a string in C++Seriously? Did you even try Googling this? The first result literally answers your question. I'm not even going to dignify this with a real answer because frankly, if you can't figure this out yourself you should probably reconsider whether programming is right for you.Also, you didn't provide:A minimal reproducible exampleYour compiler versionYour OSWhat you've already triedWhy the 12 existing answers on the duplicate question don't work for youAnd before anyone asks: No, std::string does not have a built-in split method, but I'm not going to tell you what to do about that because this is a duplicate.Voting to close. -1 for lack of research effort.(Posted 4 seconds after the question was submitted)
>>108764925Even before ChatGPT i was already browsing the documentation before SO, a lot of times I saw "question already solved" from 5 years ago and it didnt work.
>>108764925>Why did it die?Primarily due to AI psychotics (especially brown ones, as AI users tend to be) swamping every question with LLM-hallucinated answers trying farm updoots. There's no advantage to using such a platform over trusting hallucinated LLM slop when 90% of the users are brown AI-believing mongrels who post LLM slop to answer your question, you might as well go directly to the source at that point, or if you're over 90 IQ, ask in a more exclusive venue or better yet, someone competent in real life.
>>108765095The same thing is happening to all Google searches, not just SO.AI will eat itself, until there is nothing left. At that point, forums will be useful again. It's a natural process of composting.tl;dr just wait for the hype to die and it will fix itself.
>>108765127Ultimately, Stack Overflow killed itself. What an utter dogshit community.
>>108765080Maybe you are just an NPC asking only generic questions that were already answered a hundred times. The fact that you are even in this exchange unable to imagine a question that hasnt yet been asked or answered points strongly towards this being the case.
>>108765095i ask the same question about the future of coding in general (as being hyped by ai bros), what will ai train on if everything in future is ai generated?What about new frameworks, languages anything? Surely we will need new stuff as tech advances
>>108765105No it's still up and I found it incredibly useful this year. I don't know why people are complaining. Sure they're a little autistic about how the question is formatted but the high standard of curation is what separates it from Reddit.
>>108765280i think the idea is they will train AI to be more dynamic by finding new documentation to fit into its context window whenever the problem involves new stuff
>>108765299This
>>108764925You can see the plateau where the overmoderation got carried away.
I still dont know who Monica is. Now I will never need to thanks to AI.
>>108764925I find it interesting how Covid19 will be carried over as an anomaly in statistics that needs to be singled out for the rest of the century at least
>>108764925retard https://stackoverflow.com/tags
>>108765309i do wonder, there's a lot more to coding than just syntax, new paradigms patterns etc. I wonder how will ai reason everything out just from docs. I guess with human assistance it could work, still seems like there'd be much less inputs compared to what we have today.At least right now all models struggle a lot with stuff that has less code on public codebases/github
>>108765088>>108765242Not an argument. I hereby accept your concessions.
>>108765127>Why the 12 existing answers on the duplicate question don't work for youYou forgot my favorite part, none of the 12 duplicates are related at all
>>108764925I've visited it lots of times but never once asked a question thereAs far as I'm concerned it's always been dead, but as long as it stays up that's fine
>>108764925right there in ur graph
>>108764937Honestly, you will learn more from obnoxious pricks on SO than LLM that tend to over gratify the user.
>>108764938INSUFFICIENT DATA FOR A MEANINGFUL ANSWER
>>108765370>all models struggle a lot with stuff that has less code on public codebases/githubI think so too.Just the other day I was using some niche .NET library from GitHub that has not much documentation, and Claude was hallucinating shit and API calls that looked like the perfect solution, if they actually existed.
because they went from this...
>>108764925>Why did it die?Use the search function :^)>[Closed]
....to this.....
....to this
>>108764938This. Everything I needed to look up there was years old already
>>108765095>>108765280AI won't need new answers. From now until the end of time, all new software projects will be written in react+typescript, the LLM's language of choice. New features will slowly stop being added to these languages and frameworks too, since that's something only humans care about, so there's no need to worry about AI falling behind either. We will simply enter a period of indefinite stasis, a sort of managed decline overseen by bots.
>>108765442I don't believe that. you have to explain in detail WHY each is not related. post it in the next 6 hours or we will delete the question. this question is locked and I will not answer it until you provide all of this.
>>108765515>you will learn more from obnoxious pricks on SO than LLMno.not in a million years.
>>108765127Don't forget>Why are you trying to split a string? Nobody splits strings, you should do _____ instead.
>>108766834The most frustrating response. Answering a question by stating "why do you want to do that. You shouldn't do that"
It tumbled over too many copy/pasted from stack overflow jokes
>>108764925a variety of reasons>>108764937ai was the nail in the coffin but it was declining well before that. my assumption is that sites like reddit were killing it too
>>108769236and then you see a question about a guy asking how to do something on the metaphorical level as shoving hot metal rods into your pee hole and he gets four paragraphs of instructions
>>108765095>Where do chatbots get their new answers from now?From the massive amount of data they get from the people asking questions directly and pushing back on answers they don't like
>>108765309that's where simulation comes in.if it can sandbox something and be given parameters, it can create a guess that could be accurate than based on hearsay.
stackoverflow makes asking questions as miserable as realistically possible
>>108769630So, they're going to get their info from people who by definition don't know what they're talking about?
>go on SO>ask question how to do X, because I know I need to do X, and not Y>some neckbeard faggot shows up saying I should do Y instead and refuses to explain how to do XI mean, sure, I could spend a bunch of time explaining the constraints I have, why I can't do Y, why I know X is the only option, but honestly, that's bullshit, I shouldn't have to explain myself. If you don't want to answer the question then just shut the fuck up and fuck off. You don't get smart boy points for arguing with the person who asked.>how dare you complain about free help???The neckbeard isn't volunteering an answer out of charity, he's doing it because he gets an ego boost.
>>108769851These same people chats can theoretically provide good feedback for training using the positive confirmations/thanks on the solutions that work.Also data from codebases and git repos that come from agentic coding usage can provide updated knowledge on working code/architecture.
>>108769851Yeah basically, you can turn that conversation into training data.>hey gpt how do I do x>great question here's how...>that didn't work>you're absolutely right! it turns out...>i'm seeing this error with the thing you said>great catch! blah blah blahThis is why the AI tools all ask for constant feedback on the response quality.
>>108764938no. ai should have brought a tons of questions like all the new things did before. not only stackoverflow died but also all the other stack websites like maths, super user, blender, gaming ... they all died around the same time.
>>108764937tbf clankers do the same but it isnt an open forum where you get publicly hazed
>users will have to verify their identity as an individual human just to post on forums so they don't get swamped by bots.This future sucks, I wish tech stopped developing past 2005.
>>108765127This complete intolerance for retards is why I like it.
>>108770050It's a shame because the other stack sites were pretty good. I learned a lot just browsing math.
>>108764925People are okay with mostly okay quicker answers from llm. Just like people are okay with faster somewhat okay art for cheaper or free.
>>108764925midwits replaced
>>108764925
For me, it was ExpertSexChange.com.
>>108770594The answers from Stack Overflow are often not better and sometimes even worse than answers from LLMs.
>>108764925>Wah my post got deleted.lmao. You should have been looking at the windows tag before chatgpt. Shit was pretty much just.>Halp i got virus wat do? Or other non programming questions. Which get deleted>Retard c casting a std::string into a wide variant winapi function or other various mojibake issues.>Someone trying to use PostMessage to write a bot and complaining it doesn't work, half the time with a python tag as well.>Idiot calling GetLastError way to late >>108769899>explaining the constraints I haveOne of the most important things you need to do you retard. Like if the context is you are writing a driver, then no shit you cant use Y. X might also be the stupid way but there also could be a Z way which is lesser known due to the context.
>>108772071nta, but double niggers like you is why every (((community))) like this went to shit. I don't care there exists Z method, or that you dont know my context. I asked a question about X, not what's the best method, so post answer about X or shut the fuck up, you ginormous faggot.
>>108764937
>>108765317One of the things I noticed whenever SO came up in my searches is that most of the useful responses are from 10+ years ago, and most of the useless responses are from the last decade.There questions seemed equally detailed and pointed, the responses just got worse.It was already dying by the time LLMs became viable. Sure LLMs are unreliable as fuck, but as long as it gives a valid answer within a few tries more often than the droves of triggered SO responses, people will prefer it.
>>108771177
>>108772621>>108771177>have some BS project with work>decide to ask question on how to do this thing>detail out budget is 0 and I gotta use X>reply 1: LOL why would you need to do that don't do that use Y>updoots out the ass>Thanks but no I can't use Y this is a specific niche thing, I've seen others do it I know what the issue is I just am not sure on the config error I have. I know it's on this OS but due to how the system is built I can only do it this way>FURRYMASTER69: HEY YOU SHOULDN'T DISMISS A SOLUTION. DOING X IS DUMB AND YOU SHOULD QUIT YOUR JOB IT'S NOT POSSIBLE TO DO IT THAT WAY. JUST BUILD A NEW SYSTEM>BASED YES X IS BEST WAY 90,001 POST SAMA!>okay well I can't exactly do that due to needing this done end of week and it's wednesday already>faggotmod-mister: Yeah furrymaster is right if your boss tells you this you should do it another way I suspect this is a troll>locked>watch some random yt video>found the solution
>>108764925>this is how you do what you asked, but you should consider this instead if you can>that's dumb, why would you want to do that, everybody does this insteadEven if the suggestion is not applicable, the first answer is sustainable, the second is not. Stack Overflow leaned hard into the second type.
>>108764925One time when I was searching for an answer to a problem I had, I found two.The answer from ~2009 said there was an undocumented built-in function that did exactly what I wanted. It worked flawlessly.The answer from ~2017 said you must use some bloated thirdparty library, you shouldn't ever rely on an undocumented function. It wasn't even compatible with my version.The shift from actually answering questions to pretentiously spouting dogma is a travesty.
>>108772596>It was already dying by the time LLMs became viableTrue dat. Originally it was a good place for knowledgable people to ask a question and while waiting for an answer browse other questions and see if they could use their domain specific knowledge to fix other user's problems. But then the reddit effect happened and weirdo shutins with nothing better to do would just go around farming credits but answering as many questions as they could, using their voting pull to downrank questions that fucked with their meta by not being trivial to answer.And then the LLM slop answers came and washed away the last use case of the whole platform. Now you're better off trusting the AI slop from gpt and claude directly rather than risking it all on the laundered slop output of someone else's "answer this SO question" prompt.
>>108773328It's mostly because people are trying to answer as many questions as they can, even ones they can't. They can't answer your specific question about your specific use case, so they try and engage with the sort of related thing they sort of know how to do and call you stupid for insisting they answer your question as written.
>>108764937Or they try lecturer you on doing it the right way and downvote your post. Fucking hate those twats.
>>108764937>WHY WOULD YOU DO IT THIS WAY?
>>108773819>>108764937My favorite is "WHY WOULD YOU WANT TO DO THAT?" Followed by an obviously inferior way that nobody would ever want to use because of technical complexity, shortcomings, and limitations. Of all the things I'm concerned about AI destroying , Stackexchange is not one of them
I used to keep a list of good questions with good, detailed and accurate answers that got locked by fussy stick-up-their ass autistic mod triggered by the framing of the question.But I mostly stopped using the site and so stopped caring.Ironically, or perhaps not so ironic, the other day I got a garbage answer from an LLM so I clicked the source link and wound up on stack overflow from where it had copied the answer almost verbatim.
>>108764937AI is effectively worse BTW.
>>108772071>One of the most important things you need to do you retard. Like if the context is you are writing a driver, then no shit you cant use Y. X might also be the stupid way but there also could be a Z way which is lesser known due to the context.It is not the neckbeard's place to question why I want or need to do X, it is his place to either explain how to do X, or to shut the fuck up. If he really wants to suggest Z then he can append it to his answer on X.
>>108773498>But then the reddit effect happened and weirdo shutins with nothing better to do would just go around farming credits but answering as many questions as they could, using their voting pull to downrank questions that fucked with their meta by not being trivial to answer.This is evident when they don't do follow-ups to questions or corrections posted by OP. They clearly just want to farm for points instead of helping people.
>>108774686Thanks for confirming my theory and sorry bro, but you can't reliably simulate input with PostMessage, that's not what it's used for. You've been told this a million times, just because it works with notepad doesn't mean it it's going to work with your shitty NTR games.
>>108775374>gets replaced by an LLM that just gives the (correct) answer the neckbeard refused to
>>108771177>>108772621LLMs are this but you don't get to scroll through all the answers, even the downvoted ones
>>108765948literally me, lol
>>108764925the mods are colossal douchebags
>>108770050media ownership shadowbanning stack overflow results?
>>108769851nigger you and the other doubters are assuming there are problems in programming that arent solved, how preposterous you are a larping faggot plain and shrimple, theres literally nothing you can imagine that current LLMs couldnt adapt to, its just the way the software works
>>108770050>no. ai should have brought a tons of questions like all the new things did before.AI doesn't tell me to use the search function when I ask it a question.>waaa but you have to check the results!Yeah no shit nigger. Were you just blindly trusting the retards on Shitoverflow before we had AI?
>>108764969What would even be the point of going to stackoverflow, LLMs can literally generate the code customized to your specific requirements
>>108765095There are data companies that create and curate data for AI companies. Top AI companies have acquired AI IDE companies which can give them access to user data which helps them identify the gaps in AI generated code
>>108764925Duplicate question, see >>108768242Closing.
