So I created this code to reorganize my 5,000 GitHub starred repositories.

It took Claude Code a little over two hours to sort everything into structured categories and lists, all while respecting GitHub API rate limits.

Before that, I had a “To Sort” list with more than 600 unsorted repositories, plus dozens of manually created lists. Every time I starred a repository, I tried to place it in the most relevant list. But over time, the diversity and depth of my starred repositories outgrew the structure I had manually built. My 32 lists simply weren’t granular enough anymore.

So I decided to scrap all manually created lists and start fresh.

With Claude, I analyzed the actual data behind my starred repositories and defined a new set of list categories based on real distribution patterns instead of intuition. Once the taxonomy was defined, the automated classification process began. A bit more than two hours later, everything was sorted.

Now, when I star a new repository, I finally have a list structure that reflects the real diversity of what I collect. I can sort new stars manually from GitHub (mobile or desktop), and if needed, I can always rerun the stargazer Python script to reclassify and effectively “fact-check” existing classifications.

Since this workflow proved so effective, I decided to build a dedicated search page on my site to explore all 5,000 repositories more easily.

Here is the result:

👉 https://rmendes.net/github/starred/