Nobody Asks Stack Overflow Anymore
I realized something the other day: I haven’t asked Stack Overflow a question in months. I haven’t even searched it much. When I get stuck, I ask ChatGPT, and most of the time I get a decent answer faster, phrased for my exact situation, without being told my question is a duplicate.
I’m clearly not the only one. Reports suggest Stack Overflow’s traffic has dropped noticeably since ChatGPT came out. New questions are down. The company is scrambling to add its own AI features.
That’s a natural market shift. But it raises a question that I think the whole industry should be worried about.
Where did the answers come from?
ChatGPT is good at programming questions largely because it was trained on years of public programming knowledge: documentation, blog posts, GitHub, and - in a big way - Stack Overflow. Millions of questions, each answered by a human who had actually hit that problem, figured it out, and wrote down the solution for strangers, for free.
The AI is answering our questions with the knowledge of everyone who used to answer them.
So what happens when nobody posts the questions anymore?
The knowledge commons
Stack Overflow, for all its flaws (and it has many - the culture could be hostile to beginners), was a public commons. When someone solved a weird bug in a new library, the solution became public. The next person searched and found it. Knowledge accumulated, out in the open, in a form that anyone could read, correct, and build on.
Chatbot conversations are private. When I solve a problem with ChatGPT, the solution lives in my chat history. No one else benefits. No one can comment “this is wrong, here’s why” or “this changed in version 3”. And if the bot was wrong and I figured out the right answer, that correction goes nowhere either.
For problems that are already well-documented, that’s fine. The model already knows. But for new problems - a new framework, a new API, a bug in last month’s release - where will the answers come from, if the place where people used to work them out in public goes quiet?
This isn’t just Stack Overflow
The same thing applies to blogs, forums, documentation wikis, and mailing lists. The open web of technical writing was created by people who wrote things down because writing things down was how you helped each other and built a reputation. If fewer people do that, the shared pool of knowledge stops growing - and the models trained on it stop improving where it matters most.
I’ve been writing technical posts on this blog for close to a decade. Not because anyone pays me (nobody does), but because writing it down is how I think, and because I’ve benefited enormously from other people doing the same.
What I’m going to try to do
- Keep writing in public. When I solve something non-obvious, write a short post or a public note. Even a short one.
- Answer questions when I can. On Stack Overflow, GitHub issues, in communities. The commons only exists if people contribute to it.
- Verify, then share. If an AI gives me an answer that turns out to be wrong in an interesting way, that’s worth writing down, too.
- Support the projects I rely on. Documentation, open-source maintainers, the communities that answer questions. They’re what the whole thing runs on.
The tragedy of the commons, with a chatbot
The economics are simple and a little grim: it’s individually rational to just ask the bot, and collectively harmful if everyone does it all the time. Tragedy of the commons, in its purest form.
I’m not going to stop using ChatGPT. It’s too useful. But I want to keep putting things back into the pool, because the pool is where all of this came from in the first place.
Ask the bot. But when you learn something, write it down where others can find it.