Stack Overflow Went From 200,000 Monthly Questions To Nearly Zero (stackexchange.com)
- Reference: 0180517759
- News link: https://developers.slashdot.org/story/26/01/05/1431212/stack-overflow-went-from-200000-monthly-questions-to-nearly-zero
- Source link: https://data.stackexchange.com/stackoverflow/query/1926661#graph
Questions peaked around 2014 at roughly 200,000 per month, then began a gradual decline that accelerated dramatically after ChatGPT's November 2022 launch. By May 2025, monthly questions had fallen to early-2009 levels, and the latest data through early 2026 shows the collapse has only continued -- the line now sits near the bottom of the chart, barely registering.
The decline predates LLMs. Questions began dropping around 2014 when Stack Overflow improved moderator efficiency and closed questions more aggressively. In mid-2021, Prosus [2]acquired Stack Overflow for $1.8 billion . The founders, Jeff Atwood and Joel Spolsky, exited before the terminal decline became apparent. ChatGPT accelerated what was already underway. The chatbot answers programming questions faster, draws on Stack Overflow's own corpus for training data, and doesn't close questions for being duplicates.
[1] https://data.stackexchange.com/stackoverflow/query/1926661#graph
[2] https://developers.slashdot.org/story/21/06/02/1624256/software-developer-community-stack-overflow-sold-to-tech-giant-prosus-for-18-billion
Never understood how one was expected to contribut (Score:5, Insightful)
I attempted to answer some questions in which I happened to have some domain knowledge, but didn't have the time or the inclination to meet their requirements.
In many ways, they were like Wikipedia.
Re: Never understood how one was expected to contr (Score:4, Informative)
I had the same experience. They have a serious gatekeeping issue there. I got downvoted and I was the ONLY person answering!
Re: (Score:2)
Did you even stop to consider that your (only) answer might have been wrong?
We don't vote on an answer being better or worse than others: we vote on their objective accuracy.
Re: (Score:3)
They're worse than Wikipedia because if someone has some information to add to a Wikipedia article which doesn't have some demon editor squatting atop it like the throne of Nyarlathotep, they can just go ahead and add it right away without having to ask anyone or jump through any stupid hoops.
I'm not going to lurk and troll their stupid shitty slow site in order to find stuff that I can meet their points requirement with.
Re: Never understood how one was expected to contr (Score:3)
concur. there was some good information there, but asking and answering was a really onerous process to figure out, and i never did quite get the hang of answer vs. comment.
Re: (Score:2)
Yes, the Wikipedia similarity is because of the similarity of the contribution models. The core problem is moderation - you need rules tight enough to keep overall quality high enough that people trust it enough to want to visit. But doing that raises the cost for all contributors high enough that a lot of folks don't want to put up with it.
Kind of an idle question, but I wonder if Spolsky saw it coming or got lucky.
Re: Never understood how one was expected to contr (Score:1)
Why not allow everything and let users do their own filtering?
Re: Never understood how one was expected to cont (Score:2)
The purpose is to provide correct answers.
You smell more and more like a bot.
Re: (Score:2)
This is the answer to the question "Why was Stack Overflow more successful than forum posts?"
People have limited patience to find answers, a curated knowledge repository is more valuable than other uncurated options.
Re: (Score:2)
This is both reason for success of SO and the cause for being the perpetual punching bag. Since there was a high amount of moderation the answers were way better than sorting through forum posts. But people care about their question being answered and don't care that the added volume of duplicates makes the site less searchable for others. And for people answering getting moderated for "good enough" answers is often a stronger deterrent to contribution than fake internet points and altruism is an incentive.
Re: (comparison to Wikipedia) (Score:2)
I have a [1]nonzero reputation [stackoverflow.com] on Stack Overflow, but I haven't written an Answer since May 2024 and the last of my Answers that has more than one upvote is from 2020. I for one hope it does weather the storm, but the high bar to making a "good" contribution is a challenge.
And the comparison to Wikipedia is apt in some ways; but Wikipedia also has a lot of content on history, politics, and religion, and in the past had (perhaps still has) some editors who wanted to make sure certain modern social causes were
[1] https://stackoverflow.com/users/105474/david
Who did Stack Overflow kill back in 2014 (Score:3)
I used to give free IT Support in yahoo chat groups in the 1998-2003 era. People stopped asking for help in those html-based real-time chat forums sometime around when Stack Overflow was getting started.
I think this news is just evolution that has no "right or wrong", only "kill or be killed, eat or be eaten".
Re: (Score:2, Insightful)
My concern is that the replacement really isn't sustainable. ChatGPT is not going to get better, the concept is ultimately technically limited to the level of reliability we see today, and without people publishing answers to more current questions, it's going to get stale and lose its usefulness pretty quickly.
I still find it hard to understand why that's so difficult for generative AI boosters to grasp. How will it be more accurate in 2035 if most of the information it bases its output on comes from 2025?
Re: (Score:1)
Totally agree.
Also, where will AI gain new knowledge after authors & content-generators stop creating "novel" content, and stop applying the real intelligence that feeds the artificial one?
Re: (Score:2)
Man I long for the usenet of yore. Reddit is as good as it gets, and it is sometimes an adequate replacement.
Incidentally, AI *could* be awesome for technical questions if there were decent documentation out there for systems (plus source code for software systems) and the LLM had access to it.
Re: Who did Stack Overflow kill back in 2014 (Score:1)
And yet the number of GitHub projects that don't even mention their *purpose*, let alone proper documentation, boggles my mind
Re: (Score:2)
I recall using Experts Exchange before Stack Overflow became a thing.
And by "using" Experts Exchange, I mean a link to a post on their site would come up when I would search for something online. The answer was ostensibly paywalled, but all you had to do was view source and the answer was right there, buried in the HTML. You just had to search the document source for "accepted answer" or whatever
Good while it lasted (Score:5, Funny)
I was involved with Stackoverflow in the early days and it really was a gamechanger. It was so much better than trying to find your answer by slogging through forums, that it really took off. And gamifying the answering of questions really helped it grow fast. But there's always this point where the community gets swamped with rules-obsessed bureaucrats and drives the original contributors away, and that's what happened with me. The mods all became power-trippers obsessed with closing questions and keeping people in line, rather than the original purpose of helping people. And while Stackoverflow was in decline, LLMs are a massive nail in its coffin. Why would ask a question on Stackoverflow, only to be treated like garbage, when you can ask a chatbot the same question and have it confidently tell you a subtly wrong answer, but politely?
Re: Good while it lasted (Score:1)
If you keep asking, will the LLM keep trying to get it right, until it does?
Probably just bad questions and answers (Score:2)
From my experience, I mostly find answers in several year old postings and even that not consistently enough. There is also a lot of bad answers. Not that LLM answers are any better. There is quite a bit of slop produced both by humans and machines.
Re: (Score:2)
Tend to agree. I never asked a q on stackoverflow, but searches have ended up there. Often the question matches what I am looking for, the answers range from decent info to gibberish. I expect part of the downturn prior to AI was there were so many questions already answered, why ask again? The information repository is 90% complete, so it becomes more of a read only site. I'll still check them for answers. I'd actually put stack above AI in terms of answer quality. And as with anything online, answers are
What will AI train on? (Score:1, Interesting)
...now that stackoverflow and other types of forums are soon to be gone? I think at one point in the future, when answers to questions cannot be found online, AI will not be able to find a solution.
Re: What will AI train on? (Score:2)
I see a corollary where corporations enact mass layoffs in name of profits and efficiency, only to wonder why their customers no longer have the disposable income to buy their products.
Re: (Score:2)
Recall Henry Ford was just about the only CEO who went the other way. He dropped the price and gave his workers significant raises so they could afford to buy the product he was selling.
Then some assholes in SCOTUS handed the keys to the Dodge Brothers.
ESC (Score:1)
Oh well, I guess we're back to sending all our questions to Expert Sex Change.
Stackoverflow was already dying (Score:5, Insightful)
I've been on both sides of Stackoverflow over the years. After a while I got tired of responses consisting of "why would anyone want to do that?" Thanks for nothing.
I tried contributing. After several times of the mods deleting my detailed answers because I didn't add some tag or format something following the 37 requirements for a post, I gave up. I felt like everyone was missing the point of being helpful and just wanted to follow some rules for the sake of following rules.
Re: (Score:2)
I had a similar issue here. Another problem is that I found was the notion of the "accepted" answer. This is great in theory, but only works as a snapshot in time because languages evolve. There was no simple way to post the same question again, or pin the "accepted" answer to a language or software package version.
I often went down the path of implementation suggested only to find that the problem was solved in a more modern way after a version update. The community just couldn't keep up with the evolution
Re: Stackoverflow was already dying (Score:1)
There are no tags on answers â¦
Re: (Score:2)
Oh yes, the good old "why do you want to do that" reply, so you explain why you want to do that only to be told by someone else that you'll be better off buy a new PC or Reinstall the OS or something completely different, in the end you end up going elsewhere like Reddit etc and getting told how to do what you want to do - Great!!!
Now, it's just ask on of the suitable AI chatbots and you get your answer right back along with 99% working code
StackOverflow's Own Fault (Score:5, Insightful)
While I am not "vibe" coder I find myself increasingly using AI to get questions to answers that I used to post on SO, and getting pretty good results. Asking a question on StackOverflow has always been a last resort, because that is the way they want it. There is often as much energy expended on why a question shouldn't be answered as opposed to answering it. The mechanisms to search for an existing answer are not good enough to keep you from DEITY FORBID posting a duplicate. Also, the site is not good at dealing with stale content. The "accepted" answer is often no longer the currently correct answer. The level of snark and condescension is significant, and while I don't mind a bit of sarcasm now and then, I don't want to get in continual debates with someone on whether what I'm trying to do is worth doing.
Contrast the experience with AI tools. They do not yell at me for asking a question that could have been found via web search. I do not have to wait, I get an immediate answer. Yes, these tools are overly sycophantic, and sometimes they are wrong. But 90% of the time I will get a workable response and I can choose to have my code updated based upon the answer. In those cases where I don't get a workable result, I either post an issue on the repo (open source) or contact the vendor (closed), which isn't all that different from what I do with SO
Peer-based discussion are useful, and if sites like Stack Overflow disappear, we've lost something. SO can perhaps remain relevant by being better at facilitating deeper design and architecture discussions and leave the simple "how do I" questions to AI.
It's never been any good anyway (Score:2)
It was just a majority of low-skilled people asking simple questions and a minority of moderately knowledgeable people trying to make themselves look like gurus by answering simple questions. Real experts were few and far between.
At best, it was a source of inspiration as to what might be wrong with your code. You still had to take that inspiration and write the code yourself if you wanted anything good since most answers were either irrelevant, incorrect or highly specific to someone's use case.
It's sort o
next to worthless (Score:2)
i gave up on the site pretty quickly as most of the “answers” were either non-answers or attitude.
disagree with all the shade here (Score:4, Interesting)
IDK,
i bumped in to the occasional overly-rigid process hiccup at SO but by and large i found it to be a tremendously effective tool, both for asking and answering Qs. imo, the last fifteen years or so of software development owe an awful lot to SO. the rigid rules people are complaining about here (eg closing questions as dupes) definitely contributed to SO's effectiveness.
that said, i'm using past-tense for SO. as much as i owe SO, the many AI helpers are easier and usually better.
it does beg the obvious question of how the AIs are going to get training material for future technologies.
300 this month so far (Score:3)
The "monthly question volume has collapsed [to] about 300" appears to be an artifact of the last category being the current month, January 2026, even though we're only in the 5th day as of this posting. (Specifically, 321 as of Jan-5 @ 10:30 AM EDT). I expect the number will be in the 2,000+ range by the end of the month, matching both recent trends and a simple extrapolation from the first few days. (Which would still be the worst full month on record.)
Given that AI was trained on stack overflow (Score:2)
I'm not sure what's going to happen. I know that the billionaires who built out these AI coding tools aren't just going to let the whole system collapse so it makes me wonder seriously how they're going to train their AI in the absence of tons and tons of free access to programmers asking questions and programmers answering questions.
It doesn't really matter right now because it'll take a few years at least for that to be a problem but still.
Hard questions and toxic environment (Score:1)
I usually only go to Stack Overflow when I have legit difficult questions I've put a lot of time into. Last summer, I had some serious Python-Qt6 issues. I tried to be as detailed as possible and quickly got downvoted and the question was closed. I created a very clean, slim minimal example and the 2nd question got downvoted and said it wasn't minimal enough. What's worse, the close questions get removed and not even the original poster than see them in their history!!! . I had to fish them out of my e-mail.
Previous Decline vs Current Situation (Score:2)
Aggressive moderation is responsible for declining engagement over the past decade. Today, well, nobody would have gone there in the first place.
Previous Decline: When you're knowledgeable enough to properly ask the question and not be shut down by moderators you are already capable of figuring it out for yourself, so asking it is worthless unless you're just looking for likes or whatever. I guess I'm jaded because I have a history of initial configuration issues with weird conflicts breaking everything bef
Let's get our answers from AI... (Score:2)
"draws on Stack Overflow's own corpus for training data" -- and just where is AI going to get its training data on future issues that come up? You can't exactly feed it a manual and hope it will understand and be able to extrapolate answers, it just regurgitates what's already been posted with a certain degree of "confidence". So if SO goes, because everyone is relying on LLMs, it becomes and oroborous situation.
Killed the goose that laid the golden egg (Score:2)
Have LLMs killed the goose that laid the golden egg? They need to be trained on data from sites like SO, but if no one is asking new questions, there won't be any new answers, and so no new training data will become available. And if all the traffic to SO is training bots, they won't make any money from ad revenue, and so they're toast.
Re: (Score:2)
I think this might be one of the reasons why microsoft et al want full snooping of your machine. Because they can use as training data the whole loop: your questions to the AI, and what you do with the answers.
AI replaced it for me. 130%. (Score:2)
Stack Overflow was awesome, but now I have an API and documentation expert on _everything_ about software development sitting in a chatroom giving me expert answers and advice right when I ask for it. It's like Linus Torwalds and an army of software project leads are just sitting there waiting to answer when I come with my problem. I don't even search anymore, I ask the question, add layers and "if not, then how" and "how would you do this specific thing" and buddy AI spits out an essay on how to do it and
Re: Really misleading (Score:1)
Will you please accept this pity bite?
Re: Really misleading (Score:1)
You mean the quality of the answers? I stopped after around 10 questions I asked went unanswered except for some snarky replies that usually ignored the heart of the question.