2.5 million developers used it — and why I took it down

GitHub's discovery features have never been great. If you wanted to find the most starred repositories in a given category — say, authentication libraries for Node or the best machine learning frameworks in Python — there just wasn't a clean way to do it. Browsing by topic was clunky, search was useless for exploration, and comparing options meant opening twenty tabs. So I built Open Source Libs to fix that.

The idea was straightforward: scrape the entire GitHub ecosystem, organize everything by topic, and give developers a fast way to browse and compare projects side by side. I wrote scrapers that pulled repository data in bulk, ran it through a processing pipeline that extracted topics, star counts, and full READMEs, then loaded everything into a hybrid data layer — PlanetScale for the repo metadata, a file-based JSON cache for the 9,000+ topic pages that needed to render fast without hitting the database. Each project page displayed the full README inline so you could evaluate a library without ever leaving the site. Nuxt 3 handled server-side rendering, which meant every topic and repo page was individually indexable — important when you're generating a hundred thousand pages.

It took off. The site organized over 100,000 repositories across every category you can think of, and over 2.5 million developers ended up using it to research packages for their projects. It was part of a broader SEO-driven campaign I was running in 2020 alongside a few other developer tools, and it brought in solid ad revenue on autopilot.

But there was a problem I couldn't ignore. Because the site cached READMEs at scrape time, those snapshots would go stale. A library might update its API, fix a critical vulnerability, or deprecate an entire approach — and developers visiting my site would still see the old documentation. If someone followed an outdated installation guide or used a deprecated method because my cached version hadn't refreshed, that's real harm. Developer documentation isn't content you can afford to serve stale.

So I decided to take it down. The actual GitHub repositories should be the canonical source of truth in search results, not a secondary aggregator running weeks behind. The site was making money and had plenty of traffic, but keeping it running meant accepting that some percentage of developers would get bad information from me. That math didn't work.

I still think GitHub's discovery is lacking — the problem I was solving was real, and honestly still is. But the way I'd built it had a shelf life, and I'd rather kill something than let it quietly become a source of confusion for the people it was supposed to help.