Is Stack Overflow's decline really due to AI?
/ 3 min read /
Table of Contents 目录
Hi everyone, I’m luckySnail. As a programmer, back when AI didn’t exist, the site I had to open every day was probably Stack Overflow! No matter what problem you ran into, you could find an answer there; even if you couldn’t find one, you could ask the question and have elites from around the world help you out. But this site, essential for programmers, now has monthly active users far below its peak.
Is AI really the reason for this? I’ve read some community articles and discussions, and found that even without AI, it might have been on the decline anyway!
The Root Causes
Overly Strict Rules
- Unfriendly to question askers: The moderators close a large number of new questions as “duplicates” or “off-topic”, leaving the askers unable to solve their problems.
- Unfriendly to answerers as well: High-quality answers contributed by users are often deleted due to formatting nitpicks. For example, updating an outdated parameter in an old answer can get the edit rejected by the system.
Outdated Mechanisms
- The “accepted answer” mechanism is ossified: A highly upvoted answer from five years ago may be obsolete (e.g., old framework syntax), but there’s no way to automatically demote it, so newer correct answers can’t get seen.
- The platform enforced an “allow AI-generated content” policy without consulting the community, stripping moderators of the ability to delete or edit low-quality AI answers. This move is criticized as sacrificing content quality for traffic.
Beyond the above, the most important factor is the platform’s elitism. Newbies who ask questions get looked down on, turning the site into a hangout for veterans. If a community doesn’t get fresh blood, it’s not far from dying out.
In truth, the lifecycle of any community is probably something like this: Startup Phase => Maturity Phase => Decline Phase => Death or Rebirth Phase.
Of course, AI is definitely the catalyst that pushed Stack Overflow into decline. Programmers can get the answers they want much faster through AI, so there’s really no need to search and ask on the community – after all, no one likes being yelled at and getting nothing useful in return.
The Shift in Information Consumption
In the pre-AI era, searching was the main way programmers gained knowledge and solved problems. But with AI, we can just ask AI directly. If the AI is connected to an RAG system, it can perform a web search, organize the information, and answer us. Compared to traditional information retrieval, AI greatly improves efficiency and saves time. However, that doesn’t mean communities or other information channels will be completely replaced by AI. I think AI will actually force these communities to level up and do what AI can’t do – for example, pass on the energy of mutual help, solve extremely difficult problems, and build higher quality, real-time content.
For an ordinary coder like me, AI lets me get useful information faster, and at the same time makes me realize how incredibly easy knowledge acquisition has become. That forces me to think: if AI is really that powerful, what am I better at than AI? It can’t just be that I’m cheaper, right! So far, what I see is that even though AI is very capable, it still can’t truly take on product development in a real company. The main issues are:
- Team communication and collaboration – you know, dealing with people is a profound skill.
- Translating requirements into corresponding code while ensuring the functionality works perfectly.
- Designing an architecture that matches the product’s future development trajectory and keeps the code maintainable.
- Creating things that have never existed before.