17

I've been noticing stories about how our rate of questions is in significant decline, with an accompanying chart. Specifically, this MSO post from May links to a blog post about how StackOverflow is supposedly almost dead.

If we add the developments of recent months, using this SEDE query, here's what things look like: enter image description here

I dropped July 2008 and Jan 2026 to only show full month stats. After posting, I noticed links to a blog post regarding our state of affairs.

So, the decline in overall posts has continued over the past few months as well. Let's zoom in to 2025, and break the posts down by questions vs answers:

enter image description here

Has this continued trend - which has now gotten us down further than the first full month of statistics - influenced your thoughts on our state of affairs?

PS:

  • At first, this post focused on just putting the chart out there. I've re-edited it a few of times; sorry for that.
  • The chart shows all posts, not just questions (so, answers and wiki edits too).
47
  • 7
    Use a better query, no need for Excel: data.stackexchange.com/stackoverflow/query/1925371/… Commented Jan 11 at 9:46
  • 3
    related: Do you agree with Gergely that "Stack Overflow is almost dead"?, though I think new discussion is valid, even of essentially the same question there, since time has passed, and the recent numbers are much worse than they were at that point. though... the middle part of my answer there is pretty much the same for me today Commented Jan 11 at 9:54
  • 6
    FYI, you should really be using PostsWithDeleted to get the post counts for stuff like this. Commented Jan 11 at 13:42
  • 21
    What's most relevant here is that there's almost a 50% reduction in traffic from 2017 to 2022 before the AI hype started. So those ranting about ChatGPT killing SO clearly haven't gotten the full picture - if anything the AI hype just acted as a catalyst. The decline started even before the "Monica debacle" and Prosus sale (autumn 2019) - there was already a ~20% decline between 2017 and 2019. So the problem likely goes down to the core of the site: the bad moderation system and the rotten culture. So adding AI-something-something on the site ain't gonna solve jack. Commented Jan 12 at 10:10
  • 7
    @lundin That's true if you are looking at the number of new questions, but what Stack Overflow was always about is finding answers through Google search. LLM killed that. People come to read answers on Stack Overflow much less frequently. Commented Jan 12 at 11:35
  • 3
    @Dharman That would require a separate graph then, about the number of visitors. Though naturally there's a relation between the number of active posters and the number of visitors. The less posts, the less visitors. Commented Jan 12 at 11:51
  • 6
    @Lundin hard disagree on the reasons you cite. Back in 2017 we were on welcome wagon iteration #2 or #3, accompanied by the rise and quick fall of "Team DAG" in the context of which I wrote this answer. We lost the "simple" CV reasons used to close many of the bottom tier questions a few years before, and part of the earlier rise was due to SO being absolutely flooded with crap. SE overcorrected way too hard against perceived elitism into "welcoming" lots of trash, and a reduction from 2017 levels does not neccessarily feel like a bad thing. Commented Jan 12 at 12:19
  • 17
    @Lundin you don't get a site full of high quality posts on the public public internet without a good amount of "elitism" or "gatekeeping" or however you want to call it - something needs to keep out the trash instead of allowing it in. Your suggestion of giving no attention to undeserving questions seems like fantasy thinking to me - how is that supposed to work, the site magically knows which posts are unworthy and never shows them to anyone? Because a post that is displayed on my screen already uses up some attention, as does any way of marking it as something I don't want to see. Commented Jan 12 at 14:30
  • 5
    @Lundin and re explaining fluctiations - what I'm saying is, SE post numbers were significantly higher in 2017 than they should have been because most of it was bad questions. Thus a correction downwards from that level does not have to be a bad thing. And an overload of bad questions in turn explains why experts leave or don't want to stick around. Commented Jan 12 at 14:33
  • 10
    @Lundin: "There are too many (user) moderators and they are too fanatically pedantic." - is that an assumption (given the drop in posts), personal impression (given your experience), or the result of some kind of statistical analysis? Commented Jan 12 at 14:42
  • 3
    @Lundin "don't give a damn"/"let it die naturally" works if you have a decent amount of decent questions and a low amount of crap questions. It does not work in e.g. the python and java tags where you have a high amount of crap and a small amount of decent questions (see my comment below the answer by Thom for recent examples). My version of not giving a damn is that I will simply cast a DV and CV and by this hopefully help other people who no longer have to see that post. And IMO saying close voters are too pedantic and have too little domain knowledge is outlandish, at least in my tags. Commented Jan 12 at 16:31
  • 2
    please use chat instead of comments for extended discussion, and answer posts instead of comments for answering to the discussion prompt. Commented Jan 12 at 17:46
  • 7
    @Lundin "And another realization is that SO originally was "a place for programming professionals and enthusiasts". If the intention was to build a library of knowledge, which is some manner of hi-jacked narrative of what some users think the site should be about, then Q&A is a pretty horrible format for that - always was. There's already Wikipedia which is a way better platform for such." I'm not sure it's other users who are confused. You may want to review blog.codinghorror.com/introducing-stackoverflow-com and joelonsoftware.com/2008/09/15/stack-overflow-launches Commented Jan 13 at 20:59
  • 4
    @Lundin Specifically, "Stackoverflow is sort of like the anti-experts-exchange [...] meets wikipedia meets programming reddit. It is by programmers, for programmers, with the ultimate intent of collectively increasing the sum total of good programming knowledge in the world." and "Every question in Stack Overflow is like the Wikipedia article for some extremely narrow, specific programming question." respectively. The founders, both of them, always viewed Stack Overflow questions very much like Wikipedia articles. Commented Jan 13 at 20:59
  • 3
    @Lundin I don't think anyone here thinks the site design and culture make perfect sense. At most, users here just disagree on how to change it for the better. Commented Jan 14 at 16:31

6 Answers 6

Reset to default
28

I wouldn't say that the decline in posts has directly influenced my thoughts on the site's current state of affairs. It's almost certain that people are using generative AI tools to ask their questions, and if those people want to ask those tools those questions, and (quite often) blindly trust their words, rather than those of an expert, that is their choice.

Many of the questions that were asked on the site were repetitive. I recall closing many questions in 2021 onwards as duplicates. That those questions aren't being asked any more isn't a problem. Where do you think these Generative AI tools got (stole) their words from to create their sentences? It's places like Stack Overflow that provided that content for them.

Instead, now, I see a much slower stream of questions in the tags I watch, however these are often more complex and complete. Great!

What the current state of affairs has impacted is Stack Inc's opinion on the current state of affairs, and as such they are trying different things. Part of the problem with these is that these "things" are often not what the long(er) term members of the community want, and in some cases are detrimental to both their experience and the new users Stack Inc is trying to get to participate in. An excellent example of this is the open-ended questions feature, which was released and feels like it hasn't seen a single update based on community feedback (I'm aware of one, but that was actually implemented incorrectly and still broken), and is basically just on hold until they decide what they are actually want to do with the feature, but it's been left live (and broken) until then.

Stack Inc's actions, however, do change my opinion on the current state of affairs; as a community user, curator, and moderator, these half-baked (at best), and (often) disruptive releases can be a large drain on resource of the volunteers that Stack Inc rely on.

So, for me, the TL;DR is: No, the trend of new posts going down hasn't impacted my opinion on current state of affairs directly; it's affected Stack Inc's opinion on the state of affairs, and their actions have affected my opinion.

9
  • 7
    "Where do you think these... tools got (stole) their words from" <- indeed, very often from here. In fact, I have already experienced multiple cases of someone asking me a question - about CMake or CUDA; not liking my answer; going to some AI chatbot; showing me that the chatbot gave the answer I gave; and me saying "that sounds familiar" - because it's the text from the answer to my own question, or the answer I myself gave, on SO a while back. Commented Jan 12 at 10:52
  • 1
    somewhat similar post I made a while back: meta.stackexchange.com/a/384378/997587 Commented Jan 12 at 11:16
  • 6
    Full ack. It's not exactly new that SE does things which are detrimental to SO in some way, however I feel it has accelerated over the last few years. My gut feeling is that it's caused partially due to experienced people being replaced with new people who feel the need to re-discover the reasons why certain things and rules exist the hard way instead of listening to the community, and partially due to increasing pressure to find $1.8 billion plus interest between the couch cushions. The current trend feels like the experiments will continue until there's nobody left to complain anymore... Commented Jan 12 at 12:06
  • 2
    Also, nice to hear that in some corners of SO the reduced questions have a positive effect on overall quality. I'd love that to be the case in my tags, but python and java are still full of LQ posts like this or this or this or this or this. Commented Jan 12 at 12:10
  • @l4mpi: Looked at the first example. Wow... how very entitled... it's like the poster thinks we've sold him a bill of goods and has come to demand service; and that's just the tone, ignore the lack of focus (which is a more, let's say, acceptable newbie mistake even if it merits a closure.) Commented Jan 12 at 20:29
  • " rather than those of an expert, that is their choice." - experts, or at least of a human being. Probably more the latter than the former, nowadays. Commented Jan 13 at 13:29
  • 5
    @einpoklum it didn't register as entitled for me. They didn't include salutations or thanks, so they probably understand that isn't wanted on SO; the curtness may just be attempting to formulate the post without fluff which then comes across as entitled for you. However the content is way too confused, from "what is bytecode" and "why do we need CPython" it feels to me like OP is in way over their head and needs to learn several basics (e.g. how a CPU works and what kinds of operations it can perform, differences between ASM and high level languages) which makes their post unsuitable for SO. Commented Jan 13 at 15:17
  • 1
    @l4mpi Exactly. This is a very good conversation to be had with your friendly neighborhood LLM who is a much better medium for this type of question than SO. We should almost have a "Ask an LLM?" button available, and the question being hidden if the user takes that route and gets their answer there. The user doesn't need the aggravation of the -12 and closure, but the community doesn't need the aggravation of the question, even if the user meant well. It's just a fundamental mismatch. Commented Jan 22 at 22:36
  • "if those people want to ask those tools those questions, and (quite often) blindly trust their words, rather than those of an expert, that is their choice." It's all cycles. Stackoverflow was widely criticized for encouraging a dumb copy-paste culture in the beginning. Now we are criticizing people for blindly using LLM's. Always idealizing status quo and scorning the new thing. Ask someone 20 years ago how much we should trust a random bunch of people on the internet playing around with imaginary internet points. The 'experts'. Commented 18 hours ago
13

I am going to take a contrarian position, though I understand the concern.

A while back I was more busy answering questions, for Python and had set up the following query.

[python] score:1 hasaccepted:no closed:no

(and a bunch of removals of Python libs that gets a lot of interest, but don't concern me in the least. -[libXXX] -[libXYZ])

Notice the score:1. It was there because the site was overwhelmed with questions of dubious quality from rep. 1 first time users. That was happening both due to SO's increasing popularity and the Python language now attracting so many beginners (which is a good thing and which the Python community, rightly, treasures). I suspect [javascript] would have had the same problem.

In fact, if you were to look at the incoming [Python] questions just a short while back, without the score:1 filter...

Run this query - without the score filter - and you'd be overwhelmed with a question every minute or so, typically not great quality (not just lacking in styling) from rep-1 users.

Or you know, good questions, but duplicates, from a new user who didn't really understand how to search effectively before asking.

And if you look at the historical user rankings, you will find that a huge proportion of the site's users has been rep. 1 users who only ever posted one question. If someone answered, likely as not it was never accepted. Never upvoted either since I assume few users bothered even looking (and because many are loath to upvote another answer when they've answered a question, preferring to maximize their answer's ranking). I actively tried to find a query criteria to filter on user reputation, never found it, so eventually I settled on eyeballing rep.50+ and ignoring the rest.

The score filter served my purposes, but it wasn't healthy. If everyone applied that logic, all the questions would languish.

Run the same query now and you can see a question every hour or so. With the 1-rep users being about half the volume, not 70-80+% of it.

Now, we do have a well-deserved reputation for being elitist and snobby at times. Some of these rep 1 users that never came were indeed ill-treated and had not learned to grow a thick skin yet. That is something we really ought to address. I have had it happen to me and you learn to ignore, even when it comes from a moderator (most are OK, not all).

(We should nurture and assist new users, since the volume is much more manageable)

Many, many, of these questions were just noise. I don't miss them.

Maybe the site's owners do, since eyeballs reign supreme, but that's not my concern. I don't care to answer someone's homework questions. Or something trivially basic about Python. That's a large part of why I stopped answering questions on the incoming stuff. Most of my answers nowadays tend to be on older questions that I find when researching stuff for my own concerns.

Some of those questions are just not a great fit here, no matter the hoops that get jumped through. And from the high proportion of one-question-never-came-back users, I question how much value SO was getting from them.

ChatGPT is... not bad.

If ChatGPT answers their very basic questions, good for them. And, to be honest, they are probably better off from it than when they were asking here and not getting answers (or, getting closed as duplicates, which is an unpleasant first time experience).

I am sitting on the fence regarding current LLMs. They are neither very clever, nor very stupid. For many of the over-simplistic / duplicate questions that used to come in, ChatGPT or another coding LLMs are quite capable of providing OK answers. Maybe it will make some mistakes. But in my own experience asking simple coding questions about languages or topics I lack familiarity in mostly gives good-enough results: we are not helping these people for those questions. Even if users here were patiently and diligently answering them, the turnaround time is 100x that of an LLM.

Future?

Maybe SO needs to position itself to be a second-level Q&A site, serving users after their first, LLM, pass. In turn, reselling content to LLM providers might of more value, given a higher wheat-to-chaff proportion. If the site could function at lower volume levels, why can't it do so now, with possibly better content and with better name recognition? Possibly it might require fewer staff and that is regrettable, but not the end of the site itself.

SO's value proposition is as the place for Q&A with humans. Not competing with LLMs, even if volume goes down. Competitors are Reddit and Quora and fragmentation and competition is to be watched in that space. Comparisons with dinosaur moments like BlackBerry vs iPhone are misplaced. Not even Kodak vs digital cameras which seems like a more appropriate metaphor, but is also wanting. SO "just" needs to accept that its volumes will go down and learn to position itself judiciously in the new world. And, yes, that probably means not flailing about and finding sustainable ways to monetize the site.

And finding ways to be a lot more approachable to new users. By helping those asking good questions. By being kind to those asking bad questions in good faith.

BTW In the comments below, I've mentioned YouTube several times, and more precisely, reading video comments to get outsiders' perspective. This one is recent, shows this graph and is as good as any: The death of StackOverflow. Don't know the channel, not promoting it, disagree with his interpretation of duplicates being a "personal affront / RTFM". Disagree with a lot, but it is illustrative of a lot of sentiment out there and that sentiment is not necessarily wrong.

23
  • 1
    1. Can you please post a SEDE link? It may be interesting to tweak your question, e.g. by choosing other tags. So, you're focusing on Python. You mention that rep-1 single-question user volume has dropped from, say, 80% to 50%. Well, that's nice, I suppose, but - we are talking about a volume drop of 100x from 2014 to Dec 2025 and by 30x or so over the last 2 years. That means we've lost the vast majority of the users outside that problematic category. Commented Jan 22 at 21:51
  • 3
    I assume SEDE is a query engine for SE. Don't spend much time here, so no, I have no clue. I am just giving you some feedback from the perspective of a mid-level, experienced Python user, who really doesn't fancy the good old days. Those proportions I gave are very, very, fuzzy in their quality and backed only by "gut feeling", but that is what I recall from spending plenty of time staring at the questions feed. Commented Jan 22 at 21:54
  • 3
    Also, off topic slightly, I don't "elitist and snobby" is quite it, it's not really snobbery but rather what newbie users experience as abuse - and very often silent abuse. e.g. a newbie asks a poor but salvageable question, and rather than getting suggesting "you should do X to make your question more approachable", they get a bunch of downvotes and a derogatory comment - not so much a self-aggrandizing one. Commented Jan 22 at 21:54
  • SEDE is a query engine for SE <- data.stackexchange.com ; I linked there from the question. Commented Jan 22 at 21:55
  • 2
    There have been a few YouTube vids posted about this graph. Look at their comments. "Elitist and snobby" are the kinder versions of what you can find there. Sometimes incorrectly, "duplicate", is seen by at least one YouTuber as a personal affront. "They basically told to me to read the effin manual, bro!" Commented Jan 22 at 21:56
  • 4
    Let's suppose 90% of these questions are better asked from the LLM (although - I don't actually agree that is the case, for many of the SO tags). That leaves you with 10%. We have lost 90% of those 10%. And if the trend continues, we will lose even more. Also, look at the answers: Those 90% of the questions, they weren't getting 90% of the answers, they were getting lots and lots of closures. But the answers are decreasing with decent correspondence to the questions. It's true that the answers-to-questions ratio has changed significantly, but still. Commented Jan 22 at 22:50
  • 4
    I don't really care to debate all your objections here. Giving you my answer as to why I don't necessarily see the changes as that negative, from the PoV of a, possibly unrepresentative, user. Not a staff member. And certainly not an investor. "Chasing the numbers" seems inherently silly here. Asking novel programming questions is going to be progressively harder, as more and more have already been answered. Commented Jan 22 at 23:00
  • 1
    @PhilipCouling OK, so how many users doing the answering truly miss those questions? Look at the top voted answer, basically seems to be saying the same thing. Look at the score of the question. Commented Jan 22 at 23:04
  • 3
    @PhilipCouling Do you think it was a good experience asking that flood of questions and not getting answers? Getting closed as dupes? I suggest you go read some YouTube comments to get a feel for their PoV. What's your suggestion? that we keep on trying to attract a lot of questions that we then, for good reasons, have to close for being a bad fit? Or a duplicate? Commented Jan 22 at 23:06
  • 2
    "were indeed ill-treated and had not learned to grow a thick skin yet" I really hope we don't settle (and aren't settling) for this attitude :| on the rest of this post, as an aside, there's some similarity to posts I wrote a while ago- meta.stackexchange.com/a/386984/997587 / meta.stackexchange.com/a/384378/997587 (some of my thoughts have changed since then but I've not found motivation to update them) Commented Jan 23 at 3:55
  • 1
    Are you finished editing your Answer...? (I had read and UV'ed at Rev_4, I think...) It is difficult to vote on it and fairly annoying to have to re-read it again tracking all the changes if you keep editing it again and again and again... It is now already at Rev_16...! Make up your mind... Is Rev_16 now the "final Version"...? Commented Jan 23 at 11:20
  • 1
    @chivracq: I actually think it's ok that JLPeyret is doing edits. That means he's considering what people are saying, tweaking his perception, adding information... remember that votes are mostly a gamification trick; and its answers, questions and comments that really matter. Commented Jan 23 at 12:23
  • 2
    @einpoklum Yeah sure, I'm all in favour for editing and polishing, but 16 Revs in 7 or 8 hours becomes a bit annoying after, say... 10 versions, because that means re-reading more than 10 times the same Answer... And voting on 'Meta' has nothing to do with "gamification" but expressing agreement/usefulness... Commented Jan 23 at 12:31
  • 1
    @chivracq My apologies. I don't come here often and didn't realize I was being disruptive. Point well taken. Although, no one is forcing you to read the rewrites as they go. But still, point taken :-) Commented Jan 23 at 19:04
  • 1
    Yeah, don't worry, no problem... // Just know that every single Edit (to the Question or any Answer) bumps the Thread on the 'Meta' 'Recently Active Questions' page used by regulars on 'Meta'. // A few Edit rounds after posting a Question or an Answer are completely normal, upon posting and proofreading, to correct typos, improve formatting and layout, maybe add a late paragraph, but that shouldn't go on for 8 hours. // After voting on a Post, yes I'm "ethically forced" to re-read the Post and check if my Vote is still "correct". Commented Jan 23 at 19:30
9

Based on the SEDE query that was linked in Thomas Weller's comment, the last data point shows 1447 questions per month. That's 46 questions per day in 31 days. That's a bit too low for a site that wants to continue to grow or even remain a healthy, sustainable global community in perpetuity, but it's a much more appropriate number for a healthy, sustainable community than most of the other months the site has been up.

The peak rate was ~200,000 questions in a single month... that's ~6,500 questions every day, for a month! Even if the site were only up for those three months where it was >200k per month, that would still leave us with over 600,000 questions, which I would argue is quite a healthy amount to have over the entire lifetime of a Q&A site... assuming they are all good questions.

Now, I think we'd be lucky to select 600,000 good questions on the site that are worth keeping around. There are maybe twice that many on the side that are helpful to more than one or two people. That's just my gut feeling after over a decade of using and curating the site.

To really answer your question, I think we can't just look at the decline or even the entire graph since late 2008 when the site launched... we need to first ask "what is a healthy number of questions asked per month, given the amount of people we have who can curate and answer the incoming questions?" Given the smaller user base today, it's probably not much more than a few thousand questions per month... maybe 10,000 questions per month would be a good target/maximum? Are there enough active, knowledgeable users here to answer ~322 questions per day across all the various programming languages or tools being used/asked about in 2026? Maybe... I'm not sure. But I am sure that the site would have nearly as big a problem in the other direction if we were still getting 100k-200k new questions per month.

In the context of that question, I'd say that the current posting traffic is low and cause of concern, but not horribly so. As I've mentioned elsewhere, there's still about a century's worth of backlogged questions to curate even if we completely stopped receiving any new questions today, just based on the small curator footprint around today. It's certainly not a great trend, and I think the SE staff/company ownership are not doing us or themselves any favors with how they are running the network or the company. But the current state of SO in terms of questions asked per month is, refreshingly, tolerable the last few months for the first time since the first 12 months it was around.

And, to that point, I echo what others have said: yes, there are way fewer questions here today, but the ones that do get posted seem to be higher quality/have more effort put in, on average (yes, there are still many questions asked per day that are bad or low quality/effort, but I think the average quality has risen a bit). This is helped a small amount by the Staging Ground (if only SE staff would remember that was a thing), and helped a lot more by people who don't want to put a lot of effort in going to generative AI tools to ask their stupid or low effort questions there, instead.

9
  • As if it's not bad enough that questions on SO are often met with a slew of downvotes rather than suggestions for improvement, you're now suggesting we get rid of questions (and answers?) not because there's anything wrong with them, just because they're not about popular subjects. On another note: How do we know that the base of users willing to do curation is so small as not to be able to handle thousands of questions for month? And anohter question: If there are less questions, should we not focus on encouraging people to use their time for curating older ones? Commented Jan 14 at 18:28
  • 4
    @einpoklum Where does this answer suggest getting rid of questions just because they aren't popular? Maybe I'm missing something but I don't see it. Commented Jan 14 at 20:00
  • @einpoklum John's already address the first part of your comment above. As for the second, I know this because I have been at the forefront of curation for a long time and I know it is very small group, relative to the size of the site. Re: your third question: yes, to an extent. It's still good to allow and encourage new questions and answers, so long as they're high quality. The issue with bygone eras is that there was little to nothing done to ensure new questions and answers were high quality, rather opting for high quantity instead. Commented Jan 14 at 21:14
  • @JohnMontgomery: In paragraph 3. Commented Jan 14 at 22:29
  • 2
    @einpoklum I see nothing in that paragraph that suggests removing anything at all, much less because of popularity. Commented Jan 14 at 22:48
  • @JohnMontgomery: It says that there are probably not even 600,000 good questions on the site that are "worth keeping around", and thus, all the other questions are not worth keeping around - including hundreds of thousands that are useful to a few people, again according to TylerH. Commented Jan 15 at 9:57
  • 3
    @einpoklum Suggesting that there are only X number of questions worth keeping around does not mean I am suggesting we get rid of all others because "they are not popular". There are other, more important reasons to delete questions, which almost always apply. Commented Jan 15 at 15:52
  • 1
    @TylerH : So, are you saying we should keep around the questions you said are not worthy of keeping around? Commented Jan 15 at 16:44
  • 3
    @einpoklum Yes; I'm confident there are several million questions+ of no value to anyone but the asker on the site, and likely not even to them anymore today. There's no need for the system to keep them around. There are also probably a lot (maybe not in the millions, but certainly thousands if not tens of thousands or more) of old questions that we should migrate to other sites but don't because "we don't migrate stuff older than 6 months" (as if that is an actual reason). Commented Jan 26 at 14:07
4

As a preface, I'll refer to my answer to "Do you agree with Gergely that 'Stack Overflow is almost dead'?". In particular,

For me right now, I think SO is dead when the people who are looking for what we're designed to provide (a community-built library of Q&A) no longer find us able to provide that.

I'd add now,

or are collectively no longer interested in this place as an avenue to build or find it.

For me it's hard to look at the numbers and not feel that if there's an existential moment, it's about now, or somewhere in the vicinity.

At the moment, I don't feel super strongly whether there needs to be some dramatic change here in response to the declining usage of the platform, but I think thinking about it is highly appropriate right now, and thoughts about change have been on my mind. Maybe one or two years ago I'd have been pretty staunch that things are fine how they are (like how the mechanics of the platform work, and how we use them), and that the decline isn't that big of a deal. That has changed for me; I would like to encourage us on meta to be a little more open to thinking about and having these questions and discussions.

It (the decline in platform usage) does make me feel that it's a shame- I believe this platform's goals and approach are of great value to the world. They've certainly been of great value to me.

So it makes me wonder if what this is is still felt needed by the world. If it is, what's happened? Where are people looking for their fix, if not here? But these are questions I don't have the answer to.

2

The outlook is bleak if ya asks me. The data here is sold to AI, and AI answers users' questions quickly before they even get out of their IDE or google search. (that sort of reasearch is encouraged before asking a question... as it should be). If the AI doesn't know the answer, then a user might come here to ask. Once it's been answered (which often involves getting the asker to improve their question in various ways) the AI learns that answer, so subsequent possible "hits" to that answer are circumvented by the AI very quickly. This ultimately has an effect on AI itself, as less participation at sites like these leads to less learning by the AI. Where it finds "RI" (real intelligence) and new data in the future is a big question... possibly scraping of chats, messaging, emails, recordings of conversations, input via IDEs, etc..? As is, though, SO is still serving it's purpose very well. (improving questions, providing answers, teaching AI) It's just going to have much lower traffic than in the past. Acquiring new users and keeping them active on the site may be challenging going forward. I'd expect the site to survive, just with much much less traffic than in the past... a sort of baseline pulse should keep it going. The structure of this data is very easily digested by AI, so it's certainly very useful to some very large companies. It might be helpful for them to keep a sort of buffer period on the data it consumes to help keep traffic up here. (though it may be nearly impossible to stop it from being consumed immediately)

2
  • 4
    "If the AI doesn't know the answer, then a user might come here to ask." <- Actually, the AI knows nothing... for which reason, if it wasn't trained on the answer, it is likely it will just come up with a vacuous slop answer rather than say it doesn't know. Other than that nitpick - there is the danger of the vicious cycle, which is that if not enough people hang around here, there will not be the expertise necessary to answer questions. That is, it's not just the number of answerers, which can go down with the lower number of questions, but the know-how. Commented Jan 17 at 9:54
  • A future AI may recognize a knowledge gap either by itself via introspection, via compiler failures and such, or via user feedback. In such a case, the AI (instead of the user) might also decide to ask on SO or similar, or suggest that action to the user. Commented Jan 19 at 9:01
-5

As long as there is no other online site that people would rather ask their programming questions than on SO, SO remains a best place online for asking questions (to other humans).

2025 saw plenty of continuing changes affecting global need for programming answer:

  • LLMs keep getting better
  • Devs keep getting better at using LLMs (e.g. managing context)
  • Workflows shift towards agentic coding , humans doing less thinking and tinkering

A rise of activity can maybe be expected when LLM agents themselves start to ask questions online when generation and RAG via online search did not yield results. Though there will be a lot of security and liability problems to be solved to prevent such an LLM agent to inadvertently disclose secrets.

Creating a agent tool that e.g. open source devs can use which does not just use SO in training and for RAG but also for creating new questions from being stuck might be a way to get SO number of questions up again.

4
  • 2
    I'm not actually sure workflows "shifted towards agentic coding"... but be that as it may, it's an interesting prediction - that we would start getting questions from LLMs. They would need to register, though... or do you believe that users will let LLMs generate questions to be asked on their own accounts? Commented Jan 19 at 10:20
  • 1
    If SO is the "best place online for asking questions to other humans" - then why are there so few questions then? Do we conclude that people need a lot less questions answered by humans? Commented Jan 19 at 10:22
  • Yes. Either all the people who had questions to SO in the past all went to Quora, or reddit, or whatnot. Or they all ask chatbots instead. The SEDE data does not indicate which is which, but unless you can find that great place online where people ask instead, it seems much more likely that they just don't ask anywhere online anymore. Commented Jan 19 at 15:02
  • for now, LLMs are not permitted to register or ask questions on SO (and there are reasons why it wont be allowed soon). Commented Jan 19 at 15:03

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.