Stack Overflow Seeks Realignment 'To Support the Builders of the Future in an AI World'

"The world has changed," writes Stack Overflow's blog. "Fast. Artificial intelligence is reshaping how we build, learn, and solve problems. Software development looks dramatically different than it did even a few years ago — and the pace of change is only accelerating."

And they believe their brand "at times" lost "fidelity and clarity. It's very much been always added to and not been thought of holistically. So, it's time for our brand to evolve too," they write, hoping to articulate a perspective "forged in the fires of community, powered by collaboration, shaped by AI, and driven by people."

The developer news site DevClass notes the change happens "as the number of posts to its site continues a dramatic decline thanks to AI-driven alternatives." According to a quick query on the official data explorer, the sum of questions and answers posted in April 2025 was down by over 64 percent from the same month in 2024, and plunged more than 90 percent from April 2020, when traffic was near its peak...

Although declining traffic is a sign of Stack Overflow's reduced significance in the developer community, the company's business is not equally affected so far. Stack Exchange is a business owned by investment company Prosus, and the Stack Exchange products include private versions of its site (Stack Overflow for Teams) as well as advertising and recruitment. According to the Prosus financial results, in the six months ended September 2024, Stack Overflow increased its revenue and reduced its losses. The company's search for a new direction though confirms that the fast-disappearing developer engagement with Stack Overflow poses an existential challenge to the organization.
DevClass says Stack Overflow's parent company "is casting about for new ways to provide value (and drive business) in this context..." The company has already experimented with various new services, via its Labs research department, including an AI Answer Assistant and Question Assistant, as well as a revamped jobs site in association with recruitment site Indeed, Discussions for technical debate, and extensions for GitHub Copilot, Slack, and Visual Studio Code.
From the official announcement on Stack Overflow's blog: This rebrand isn't just a fresh coat of paint. It's a realignment with our purpose: to support the builders of the future in an AI world — with clarity, speed, and humanity. It's about showing up in a way that reflects who we are today, and where we're headed tomorrow.
"We have appointed an internal steering group and we have engaged with an external expert partner in this area to help bring about the required change," notes a post in Stack Exchange's "meta" area. This isn't just about a visual update or marketing exercise — it's going to bring about a shift in how we present ourselves to the world which you will feel everywhere from the design to the copywriting, so that we can better achieve our goals and shared mission. As the emergence of AI has called into question the role of Stack Overflow and the Stack Exchange Network, one of the desired outputs of the rebrand process is to clarify our place in the world.

We've done work toward this already — our recent community AMA is an example of this — but we want to ensure that this comes across in our brand and identity as well. We want the community to be involved and have a strong voice in the process of renewing and refreshing our brand. Remember, Stack Overflow started with a public discussion about what to name it!
And another another post two months ago Stack Exchange is exploring early ideas for expanding beyond the "single lane" Q&A highway. Our goal right now is to better understand the problems, opportunities, and needs before deciding on any specific changes...

The vision is to potentially enable:

- A slower lane, with high-quality durable knowledge that takes time to create and curate, like questions and answers.

- A medium lane, for more flexible engagement, with features like Discussions or more flexible Stack Exchanges, where users can explore ideas or share opinions.

- A fast lane for quick, real-time interaction, with features like Chat that can bring the community together to discuss topics instantly.

With this in mind, we're seeking your feedback on the current state of Chat, what's most important to you, and how you see Chat fitting into the future.
In a post in Stack Exchange's "meta" area, brand design director David Longworth says the "tension mentioned between Stack Overflow and Stack Exchange" is probably the most relevant to the rebranding.

But he posted later that "There's a lot of people behind the scenes on this who care deeply about getting this right! Thank you on behalf of myself and the team."

Standing on the tracks.

[jddj]

With the freight train approaching.

Stack Overflow has perhaps more to lose from AI than the search engines, relatively speaking.

With SO you're performing a "known item search" most of the time, while "unknown item search" will be a good use case for search engines for a while yet.

Since a decent AI can plow through all the bullshit on SO, all the "what a stupid question" and "you're doing it wrong" non-answers, all the shit that worked 15 years ago but not now, it's so much better at getting to the point.

Yes, I know the AIs were trained on that content too. But do far they've been giving me a very good SNR on technical questions; far better than SO.

Re:

[Anonymous Coward]

Stack Overflow has been pretty much dead for several years already, generative AI technologies are just a final nail in the coffin.

Of the few questions still getting asked nearly all of them could be answered by a) reading the friendly manual on the specific function they're dealing with, or b) a quick Google search. In other words they amount to, "I'm lazy, do my work for me."

Re:

[machineghost]

Exactly! For years before AI, Stack Overflow has drifted from "that place devs can get help when they have a question" to "that place where every new question you ask is downvoted, (falsely) flagged as a duplicate, and ignored for days, weeks, or maybe forever ... unless you're lucky enough for a kind soul to happen upon it".

As they wrote:

It's very much been always added to and not been thought of holistically

Exactly: there was a push to improve quality on the site, which was good ... but it was made with absolutely no thought whatsoever as to the side effects. As it turned o

Ignoring the next language and API

[will4]

LLMs are diverting what once was traffic to Stack Overflow for many of the basic questions.

What happens when there is no Stack Overflow or equivalent wide-scale technology question and answer corpus for the LLMs to ingest?

For example, What will the LLMs use for training data when whatever eventually replaces JavaScript and HTML and RPC on the internet?

Re:

[dinfinity]

What will the LLMs use for training data when whatever eventually replaces JavaScript and HTML and RPC on the internet?

Newly written API documentation, all new open source code, and all the training data they already use.
In addition to that, they can also synthesize training data by (at worst) just trying out shit and effectively reverse engineering and documenting whatever interface may be of interest.

Also, note that LLMs are not the end of the line. To think that AI will somehow be fundamentally and perpetually limited by only having access to pretty much all recorded human knowledge until 2025 would be naive and shortsig

Are they concerned about "the" or "their" future?

[ffkom]

The text sounds like it tries to avoid mentioning that these companies are primarily concerned about their business model being broken by LLMs that ingested volunteer's advice as much as they did, but provide a more convenient interface to query the amassed data. That is not my concern, to me the LLMs are just consumers of free work done by others as are StackExchange et al.

The part that I find interesting is only whether there is a future where the LLMs know their limitations and are able to delegate que

Re:

[RobinH]

At least under the StackOverflow model, the content still references who contributed it, and you could theoretically point to your content and show that you were a contributor. Under the LLM model, there's zero referencing happening.

Re:

[timeOday]

Fostering a community that actively contributes questions and answers is still useful work. I don't know what the answer for Stack Overflow is because I always liked the site, but AI is their extinction-level asteroid.

Were is Wikipedia for Code

[Kisai]

Here's what I want. I want a "Wikipedia for code"
Not examples
Not idiots not understanding Javascript from JQuery
Not idiots trotting out 11 year old examples that don't compile

99% of the questions asked to stack overflow are "how do I get rid of this error", because of naïve understanding of why the error appears. A lot of programming documentation is full of mistakes or depreciated examples, so trying to figure out how to "DO THING in LANGUAGE" that works in another language and has nearly the same syntax doesn't work.

AI doesn't help here, because AI doesn't understand ANY language. But it does understand code that has to compile, so it's less likely to produce something that doesn't compile, but it won't explain why.

Re: Were is Wikipedia for Code

[jddj]

Funny. On the several projects I've tried it for, Claude always offers to explain. And when I've needed that on a particular block, it does an excellent job. It does understand what it's doing, at least to some extent.

Re:

[rknop]

Funny. On the several projects I've tried it for, Claude always offers to explain. And when I've needed that on a particular block, it does an excellent job. It does understand what it's doing, at least to some extent.

It doesn't understand anything.

It is trained to echo explanations, or at least things that sounds like explanations and may often accidentally be correct, based on all the various technical documentation you can find on the net.

Re: Were is Wikipedia for Code

[jddj]

We could have a philosophical discussion on whether knowledge requires a sentient "knower" (KM pros would tell you it does not, as that's the key to their entire domain of practice), but I'm not confusing Claude with a sentient being, conscious between questions. It is certainly not.

However, the explanations I get surpass stuff copy/pasted from docs, and indicate that distillation of knowledge has occurred.

It's sufficiently similar to "understanding" that I'll call it that for my purposes.

Re:

[dinfinity]

Define "understand".

Be precise in your definition and make sure you base it on and apply it to human behavior. What criteria do you use to determine whether someone you know understands some subject?

Then (and only then) start applying that definition objectively and scientifically to actual interactions with LLMs. If you are honest in that process, you'll come to the conclusion that it is hard to state that LLMs "don't understand anything". There are instances in which they do show significant levels of und

Re:

[Anonymous Coward]

Why would anyone want to do all that nonsense?

Re: Were is Wikipedia for Code

[NagrothAgain]

code examples on forums aren't supposed to compile, it's an example so that you can learn how to write code that will compile. If all you want is someone to write code for you, go find a library (etc.) and use it. And to be perfectly blunt, if you're not adding comments into your code saying which llm and what query you used to generate it, you're being lazy and dishonest. Because it's not your code.

Re:

[ukoda]

code examples on forums aren't supposed to compile, it's an example so that you can learn how to write code that will compile.

The problem is less code that doesn't compile but rather code that will never compile because the answer is for Python 2 or API V1.0.0 and you are wanting help with Python 3 or API 3.4.2. Of course frequent users of Stack Overflow know to be critical of answers from 15 years ago and if it's age is likely to affect what you are using, and looking for more recent answers if needed. A human with reasonable experience will soon be able to narrow it down to answers that provide the clues they were looking for.

Re:

[angel'o'sphere]

And to be perfectly blunt, if you're not adding comments into your code saying which llm and what query you used to generate it
That would be pretty pointless. As those LLMs get updated rapidly.

you're being lazy and dishonest. Because it's not your code.
Well, if it is an university/highschool assignment, then you should do that. For any other circumstances I would not know that the point is.

But you could give that as a hint to the "vibe coder IDEs", that they store the promot behind a tiny-url, and reference

Re: Were is Wikipedia for Code

[vbdasc]

Exactly. You are not supposed to ask SO to write algorithms or working code fragments for you. If you need such a crutch, then maybe programming isn't your thing (of course, with the advent of LLMs and vibe coding, algorithmic thinking is no longer a requirement for "coding", unfortunately).

I personally admit that I've asked SO questions over the years maybe a couple dozen of times, but my questions were always about how to use some under-documented feature or API, never needing answers consisting of more t

Re:

[Tony Isaac]

AI doesn't help here, because AI doesn't understand ANY language

Actually, AI *does* help here, it's great at translating code from one language to another.

But it does understand code that has to compile

That's not my experience. Almost 100% of the time, if the code generated by AI is more than trivial, I have to fix it before it will compile. It usually gets close, but it does *not* take the entire context into account.

but it won't explain why

Again, AI is actually pretty good at explaining why code works, or doesn't work.

Have you actually used an AI programming assistant?

It seems like they address this...

[SuperKendall]

Here's what I want. I want a "Wikipedia for code"
Not examples...

From the summary:

A slower lane, with high-quality durable knowledge that takes time to create and curate, like questions and answers.

Doesn't that multi-lane system they are talking about seem to address what you are after? I feel like that "slow/deep" lane would offer the kind of higher quality content you are after.

Heck they could even use AI to categorize existing SO answers into these lanes and probably get it mostly correct!

What I don't s

Re:

[angel'o'sphere]

AI doesn't help here, because AI doesn't understand ANY language.
Actually, they claim, they do.

But it does understand code that has to compile, so it's less likely to produce something that doesn't compile, but it won't explain why.
You never worked with an LLM.
Of course they explain to you why they wrote the code like that, and comment it accordingly, and explain how it works.

Or do it other way around. Copy/Paste a reasonable big chunk of code into an LLM and explicitly ask for an assessment and explanation

I don't know what they're going to do

[rsilvergun]

It's becoming an increasing problem where an llm provides a summary of content from a web page instead of anyone actually going to the web page.

It collapses their business model. Not just stack overflow dozens of pages. And that's before we talk about people using AI bullshit to spam fake slop content after it was trained on real content.

The internet is probably going to basically collapse at least a section of it that produced all the free content we've been enjoying. The AI bullshit will consume

But eventually it all collapses

[SuperKendall]

It's becoming an increasing problem where an llm provides a summary of content from a web page instead of anyone actually going to the web page.

I've been thinking about this for a while, especially in terms of places like StackOverflow.

Right now, yes it's easier to get the summary from an LLM, partly derived from places like StackOverflow.

I'm guilty of this as well, my use of StackOverflow is also way down from what it used to be.

But over time, technology grows and evolves. So what happens when there are no web pages, no StackOverflow posts to summarize? In not to far distant a future there could be a number of things where all an LLM has to go on is developer docs from the source, with no Q/A or published examples because everyone is looking at LLM output.

The only thing keeping LLMs going on longer I think is the fact that developers keep pumping out articles on various topics, but is that enough to really provide LLMs all they need to keep answering programming questions well, or even more importantly to generate whole applications from scratch?

It's a real loss I feel that there is no kind of feedback from LLMs back to our overall system. Like in the past if I were looking up something on StackOverflow and could not find a good answer (or found a better answer/modification), I would post that back to StackOverflow.

But if I tell an LLM "Thanks that suggestion worked" that feedback is lost to the void. There's no upvoting, no storage that I know of applying to all other questions people ask to let the system know "yes that answer worked particularly well". So all the LLMs can go on giving out the same answers forever not really knowing they are flawed... unless someone publishes an article about it.

Re:

[vadim_t]

But if I tell an LLM "Thanks that suggestion worked" that feedback is lost to the void. There's no upvoting, no storage that I know of applying to all other questions people ask to let the system know "yes that answer worked particularly well". So all the LLMs can go on giving out the same answers forever not really knowing they are flawed... unless someone publishes an article about it.

There's no reason why LLMs can't have feedback mechanisms. In fact they do. ChatGPT has thumbs up and thumbs down buttons.

Re:

[Paul Fernhout]

"The internet is probably going to basically collapse at least a section of it that produced all the free content we've been enjoying. The AI bullshit will consume it but then the internet will be full of AI slop and nothing else and then as things change the AI won't have anything to train itself on except more slop."

See also to support your point: "The Dark Forest and Generative AI: Proving you're a human on a web flooded with generative AI content"
https://maggieappleton.com/ai-... [maggieappleton.com]
"There's a swirl of opti

AI blended world

[misnohmer]

AI is like a search engine without sources/accreditation, so people don't actually know where the AI learned a particular coding trick, nor do they care. Plenty of new developers already never learn what an actual stack overflow is. In time they might not recognize the website name either. Stack Overflow as a company needs to find a way to stay relevant. In the short term they can capitalize on their content as training data, but at over time this training data will become stale, so the big challenge will b

They have become irrelevant...

[tonymercmobily]

Just like Microsoft Windows, or even Google...
Things that once seemed untouchable and impossible to beat, were beaten by changes in the tech landscape.
Yesterday, Toyota won with amazing engines.
Today, the engine is irrelevant, what has become relevant is the battery tech.
I myself spent considerable amounts of time on either SO or Google and thought: when are these ever gonna be repaced.
And here I am. I have barely touched either -- while still actively working -- in months and months.

They are irrelevant, an

Re:

[Tony Isaac]

These technologies you declare dead, aren't quite dead yet.

Data analytics tools like PowerBI or Tableau, and NoSQL, were supposed to make SQL databases obsolete. Guess what, SQL is still growing, not shrinking.

Windows is on more desktops then ever, and Google is evolving, but it's not going anywhere.

Stack Overflow is not in the same league as these other tools. It held a prominent place among developers, but its scope and capabilities were always quite limited. I always found it frustrating to use Stack Ove

Re: They have become irrelevant...

[nasch]

You know something like 70% of new cars have an engine, right?

Re:

[Slayer]

At least in the case of stackoverflow they were not all that innocent about their own demise. They actually sold their content [slashdot.org] to AI giants from the onset. They did not even ask the actual content creators [slashdot.org] for their permission. Instead of training their own AI on their own content they "maximized short term profit" and sold out their community.

They have it coming to them, and rarely has anyone deserved this so much.

AI/LLMs extract value

[Required Snark]

And after reaping the economic benefits they leave nothing behind.

Kind of like clear cutting a forest and not even pretending to replant. All that's left is eroded soil, scrub wasteland and ecological disaster for whoever comes next.

Blaming AI

[SmokyMarshmellow420]

The demise of Stack Overflow began a long time before AI was a thing, due to the horrible community. It was full of beginner level noise, and real useful questions went unanswered or were closed with some B.S. reason, rendering it only useful for mundane tasks mostly related to poorly engineered web development tools.

Re:

[Registered Coward v2]

The demise of Stack Overflow began a long time before AI was a thing, due to the horrible community. It was full of beginner level noise, and real useful questions went unanswered or were closed with some B.S. reason, rendering it only useful for mundane tasks mostly related to poorly engineered web development tools.

Which describes Stackoverflow, or many “communities’,” for that matter - self described experts looking down at noobs looking for help while complain there questions don’t get answered because well, no one has the expertise to answer them.

Re:

[angel'o'sphere]

It is more a problem with bullies.
If they write something wrong, and you dispute it they retaliate with: "who of us has more gold medals, you or me? Who is likely right -- cough cough, or believed?"

The one I remember claimed he had written compilers for the JVM and CLR, and did not know what a stack machine is.

And sure as hell, he got voted up and one or two of my answers/comments in that thread got voted down into oblivion.

is that so?

[znrt]

Software development looks dramatically different than it did even a few years ago

no, it doesn't. ai mostly replaced freshmen and the way early prototypes are made ... and paved the way for a horde of ceos and mbas to create the next clusterfuck in 3, 2 ...

the company's business is not equally affected so far

how much for selling user content to openai? surely a lot. not that that's going to last very long if the ai is going to have to much on his own slog ...

OpenAI : All Your Data Belong To Us

[thesjaakspoiler]

StackOVerflow just got their asses scraped, something they should have never allowed.
WebScrapers have been out there for about ... 2 decades or so.

Re:

[angel'o'sphere]

They got paid for it.

Soo, marketing bullshit overload

[gweihir]

That does not bode well...

Stackoverflow has devolved into narcissism

[hydrodog]

I used to ask lots of questions on stackoverflow. I answered quite a number at first also, but then the topics I knew were mined out. Even so, I was in the top 4% by asking great questions that would over time, get voted up. Today, I can get a lot of simpler questions answered faster with chatgpt, but when I am stuck, when I really want a human, they are always way more interested in playing gotcha and downgrading my question. They have such a narrow idea of what constitutes a valid question, and even when some people are willing to answer, I have had people delete their answers (fortunately I saw one first). The site is dominated by assholes trying to prove how smart they are. And the people running the site think they are absolutely right, so when I complained on meta, they basically agree with the way it is. This is exactly analogous to a restaurant which hires servers who are nasty and sarcastic, then acts surprised when people stop coming.

Re:

[dcollins]

I've said for quite some time that an essential problem with SO is that, since people are incited to score points, they want that to come easy, and so they get irritable when there's a fundamentally hard question that gets asked. Hence the downgrading and looking for reasons to close or delete a question. Which especially sucks for high-knowledge question-askers who have already thought through, researched, and ruled out any easy solutions.

The Iron Law of Stack Exchange [madmath.com]

100 years ago it was the "Radium ore Revigator"

[DjangoShagnasty]

Now we have Temu Skynet.

And where will AI get it's data from?

[jeffy210]

As more sites start seeing a drop off due to people getting their answers from AI, where will AI get it's data to train from if it has nothing to scrape? (Legalities and morals aside). It feels like it'll become self defeating over time or hit a wall of not having any new things to reference.

the dustbin of history

[ZipNada]

I've consulted SO quite a bit (I call it it slack overflow) and have found a lot of very helpful answers there. For coding issues I can usually rely on the AI now, it already ingested the contents of SO. For anything to do with tricky system problems, how to set something up, etc. SO is still a good resource. My number of visits to the site have definitely declined though.

I don't know what they can do to resurrect it. Inject a bunch of AI that they rent from someone else and hope it is unique and relevant? It doesn't seem likely to fly.

It's dead, Jim.

[vbdasc]

Well, Stack Overflow is finished. It might get some additional unlife by jumping on the AI bandwagon, joining the other slop factories. But what we know and used is gone, or it will be shortly.

I fear the time when the same shit will happen to Wikipedia.

Garbage in, garbage out.

[devslash0]

How on earth are they hoping to be the foundation of future AI if 99% of SO replies are completely useless or irrelevant?

Might be nice if LLMs led to communities

[mattr]

At the moment the people training LLMs basically stole the knowledge in SE/SO and packaged it better.
It took too long (i.e. not immediate) to get an answer, and the old ones have old info.
LLMs also have old info.
So LLMs need to recommend newer info over old, deprecated, or hallucinated info.
Also it was nice and useful to see people trying to help solve your problems when they were tough. I would like to see that kind of community come back and LLMs could perhaps drive you to them, i.e. link back to a page o

Well... have they not noticed...

[zkiwi34]

The AI bots have almost certainly scraped pretty much everything from such as stack overflow, GitHub and the like.