I spent a weekend digging through the top posts on Hacker News. I wasn’t looking for the next big startup. I was looking for ghosts.
You know the feeling: you see a shiny new GitHub repo, 5,000 stars, a README that promises to solve everything. You clone it, run it, and… nothing. The last commit was three months ago. Issues are piling up. The author vanished. Another AI-generated corpse added to the pile.
I analyzed the GitHub repos posted to Hacker News over the past year, focusing on how AI-assisted coding has changed the landscape. The numbers are terrifying. The number of new projects has exploded—but the number of active, maintained projects has barely budged. We are drowning in a flood of half-finished code.
AI doesn’t democratize software development. It accelerates the creation of digital litter.
Every day, thousands of developers use tools like ChatGPT and Copilot to generate entire repositories in minutes. They push them to GitHub, post them on Hacker News, and bask in the upvotes. Then they move on to the next shiny idea. The code rots. The issues go unanswered. The repo becomes a monument to the illusion of productivity.
You’ve probably seen this yourself. You’ve starred a dozen projects that looked promising, only to realize they’re all abandoned. The problem isn’t that AI can’t write good code. It can. The problem is that AI can’t care.
The bottleneck is no longer writing code. It’s caring enough to maintain it.
This is not a failure of technology. It’s a failure of incentives. The current ecosystem rewards creation—stars, followers, HN front-page glory—but never rewards maintenance. Why fix a bug in someone else’s repo when you can generate your own? Why maintain a project when you can start a new one that gets more attention? AI supercharges this cycle, turning open source into a landfill.
I’ll say it plainly: the open-source apocalypse is here. Discovering genuinely useful, well-maintained projects is becoming impossible because the signal-to-noise ratio has collapsed. We are one bad update away from a world where every new project is AI-generated, abandoned within a week, and the only thing that survives is the code that someone actually cares about.
So what do we do? Stop celebrating the sheer volume of code. Start demanding proof of maintenance. Look for projects with recent commits, active issue discussions, and a clear roadmap. If you’re a developer, resist the temptation to generate and abandon. If you’re a user, reward the boring maintainers who keep the lights on.
AI can write a million lines of code. It can’t write a single moment of dedication.
The next time you see a viral GitHub repo, ask yourself: is this a tool I can trust, or is it just another tombstone in the graveyard of good intentions?
FAQ
Q: Isn't AI-generated code still useful even if it's not maintained?
A: A one-off script or a finished tool can be useful, but most AI-generated projects are frameworks or libraries that require ongoing updates. Without maintenance, they become security risks and compatibility nightmares. Unmaintained code is worse than no code.
Q: What's the practical implication for me as a developer or tech consumer?
A: Stop trusting popularity as a signal of quality. Check the last commit date, the number of open issues, and the frequency of releases. Prioritize projects with a clear history of maintenance. If you're creating, commit to a maintenance schedule or don't publish at all.
Q: Isn't the contrarian view that AI-generated code is still a net positive because it lowers the barrier to entry?
A: Lowering the barrier to entry is great for learning and prototyping, but it's destroying the trust infrastructure of open source. The real value in software has always been in the community, documentation, and long-term care. AI gives us more code, but less of what actually matters.