You’ve felt it. You open GitHub, looking for something interesting—maybe a side project that sparks an idea, maybe a tool that solves a problem you didn’t know you had. And what do you get? A trending page that feels like it was designed by someone who has never actually used the internet.
GitHub’s discovery UX is, to put it kindly, not good. It’s a graveyard of “Explore” tabs that lead nowhere, topic pages that haven’t been updated in months, and a search bar that rewards SEO-optimized README files over actual innovation.
The irony of the world’s largest code repository is that it’s terrible at helping you find code.
So someone built their own solution. Roamers.dev is an endless feed of GitHub projects—TikTok for code. You scroll, you discover, you move on. It’s free, it’s anonymous-friendly, and it works. Sort of.
Here’s where it gets interesting. The top comment on the launch thread wasn’t praise. It was a diagnosis:
“I think it speaks about the current state of github that I had to scroll 12 projects to find the first entry not related to AI.”
Twelve projects. Twelve. Before you find something that isn’t another LLM wrapper, another RAG pipeline, another “ChatGPT for X” clone.
The discovery tool didn’t fix the problem. It just made the problem scrollable.
This is the uncomfortable truth that Roamers.dev accidentally exposes: GitHub isn’t a diverse ecosystem of innovation anymore. It’s a monoculture. And the monoculture is AI.
Now, you might think: well, AI is what’s hot right now. Of course it dominates. And you’d be right. But that’s exactly the issue. The endless feed format, borrowed from TikTok, is designed to optimize for engagement signals. And what gets engagement? Whatever’s already popular. The AI projects get clicks, so they get surfaced more, so they get more clicks. It’s a self-reinforcing feedback loop that the tool cannot—and was never designed to—correct.
Serendipity doesn’t survive optimization. The algorithm doesn’t want you to discover something new. It wants you to stay.
This isn’t really about Roamers.dev. The tool is well-built, and the intention is genuine. It’s about what happens when you take a broken discovery system and slap an infinite scroll on it. You don’t get better discovery. You get faster consumption of the same stuff.
Think about what made GitHub exciting in the first place. It wasn’t the scale—it was the serendipity. You’d stumble onto a project that solved a problem you didn’t know existed, written by someone you’d never heard of, in a language you’d barely touched. That kind of discovery requires friction. It requires a system that doesn’t just show you what’s popular, but shows you what’s weird, what’s niche, what’s quietly brilliant.
The TikTok feed model is the opposite of that. It’s designed to reduce friction to zero—to make scrolling effortless, to keep you in the loop. But zero friction means zero surprise. You see what the algorithm thinks you want to see, and the algorithm thinks you want AI.
The promise of discovery was never about seeing more. It was about seeing differently.
So here’s the real question: can we fix GitHub discovery without reproducing the monoculture? The answer isn’t another feed. It’s probably something messier—curated lists, human editors, serendipity engines that deliberately surface the unpopular. Things that don’t scale. Things that feel inefficient.
Because the alternative is what we have now: an endless scroll through a hall of mirrors, where every project looks like the last, and the only thing that changes is how tired your thumb gets.
Roamers.dev is a clever hack. But it’s also a mirror. And what it reflects isn’t pretty.
When the cure looks just like the disease, maybe the problem isn’t the medicine. Maybe it’s the patient.
FAQ
Q: Isn't AI dominating GitHub just natural market behavior?
A: Yes, and that's exactly the problem. Market dynamics optimize for what's already winning, not what's interesting. Natural market behavior is how you get monocultures. Serendipity requires deliberate friction against the grain.
Q: So should I stop using Roamers.dev?
A: No—it's a genuinely useful tool and the intention behind it is solid. Just go in with eyes open. The feed will show you what's popular, not what's surprising. Use it for trend-watching, not for discovery of the unknown.
Q: Is serendipity actually valuable, or is this just nostalgia for a GitHub that never existed?
A: It's valuable because innovation doesn't come from the mainstream. Every breakthrough project started as something weird and unpopular. If your discovery system only surfaces what's already winning, you're not discovering—you're confirming what you already know.