I am deeply concerned for the future of software development now that the AI era has completely taken over how developers use the internet to seek answers to their questions.
The "old" way was you asked a question on a search engine (Google in most cases) and the search index returned a set of links. You reviewed the results and clicked a link, visit the website, read the information provided, and if it applies you interact with the site: sign up for an account, comment, reply, up vote, and so on. This process engages you as a the developer with the community and your interaction further enhances the value of the website. As this community grows more up to date information flows in, deprecated knowledge gets updated, and the value of the community increases.
What I am seeing happening today is you ask your question on your search engine (or in your IDE or CLI) and what you get in response is an AI generated answer that in most cases is pretty good. Your answer most likely is provided directly in front of you, gathered from most likely the same sites you would have visited in the past, but now there is no need to explore further since you have the answer. No need to visit the source website. No engagement with the community, no more contributing or up voting. You have the answer and you move on.
This is great for today. What happens in 5 years? Sites like Stack Overflow, reddit, Dev.to are losing traffic to the AI scrapers and the long term effect will be a significant decrease in engagement and community building, then knowledge. I expect the AI itself will stop learning, it will be stuck in 2026 knowledge, as there will be limited new information to scrape.
I'm not sure we are building the future we want. Let's check back in a year or two and see how this is going.
Top comments (2)
Thanks for sharing this, John. I share your concern about losing the feedback loops that create and improve technical knowledge.
When developers stop participating in these communities, we lose:
Over time, this could mean less interaction between people, less up-to-date knowledge, and greater dependence on AI platforms. That worries me even more when those systems are essentially closed black boxes.
Depending on proprietary platforms such as Stack Overflow or Reddit was never ideal either, but at least the knowledge there is publicly accessible, linkable, attributable, discussable, and created through visible human interaction.
Maybe this is part of why I've recently become interested in the IndieWeb. For technical communities, I find the idea quite compelling: developers publishing under their own domains, linking to each other, discussing, responding, and collectively sharing knowledge on the open web.
Maybe that sounds a little utopian, but I'd like AI to push us in that direction rather than the opposite one 😄.
Another side of AI gives me some hope: AI can dramatically reduce the effort required to document and publish what we learn. Countless useful blog posts were probably never written simply because the developer didn't have the time. AI can help turn draft notes, experiments, debugging sessions, or lessons learned into something worth sharing.
Here is a recent example, written with AI assistance (not that technical), but about what I have recently learned about the #IndieWeb: 🌿 Second steps into the IndieWeb.
I guess the healthiest way we can use AI is not only to consume knowledge faster, but also to make it easier for us to contribute more knowledge back to the web 🤞🙂.
Terrance Tao and Gary Marcus shared the same sentiment, albeit for mathematics and AI in general, but the idea is exactly the same concern I have for software development future - garymarcus.substack.com/p/two-dire...