GenAI, the Snake Eating Its Own Tail

https://news.ycombinator.com/rss Hits: 1
Summary

Generative artificial intelligence (GenAI) tools such as ChatGPT and Claude have two superpowers. The first superpower is a boon: they can dramatically increase human productivity. I use them on a regular basis to answer questions, learn new skills, write code, create images, and much more, all at a rate of speed and quality that was science fiction just a few years ago. The second superpower is a bane: GenAI is quietly destroying the very ecosystems that made it possible in the first place. Under the hood, GenAI is built on large language models (LLMs), which are able to extract patterns, structure, and statistical relationships from massive data sets. These data sets consist primarily of content created by human beings: books, blog posts, articles, forum discussions, open source code, art, photography, and so on. LLMs are able to extract value from this content at an unprecedented scale, but all that value is captured by the GenAI company and its users. If you’re a content creator, you get nothing: no attribution, no referral traffic, no revenue share. Not even a thank-you. This feels unsustainable to me, a bit like a snake eating its own tail. In this blog post, I’ll go through three examples of how GenAI is destroying the very ecosystems it relies on, and then discuss possible solutions that may give everyone (users, GenAI companies, and content creators) more value. Example #1: online communities For many years, StackOverflow was the most popular Q&A site for programmers. Any time you hit a weird error while coding, you’d do a search on Google, and more often than not, find a good answer on StackOverflow. But now, in large part due to GenAI, StackOverflow is nearly dead: StackOverflow questions over time. Source: Marc Gravell. Although StackOverflow’s decline started before GenAI went mainstream (ChatGPT was first released in 2022), GenAI accelerated that decline considerably. That’s because nowadays, instead of searching around for an answer on a Q&A site, and...

First seen: 2026-01-21 18:40

Last seen: 2026-01-21 18:40