About the data

Where it comes from

Everything is read from GitHub's public search API. GitHub returns at most 1,000 results per search, so the fetch splits the star range into slices that each stay under that limit and stitches them together. A scheduled job does this once a day, rebuilds the site and publishes it. Nothing is scraped, and the site has no server: it is static files plus the data you see.

What a rank means

Rank is position by star count at snapshot time, nothing more. Stars measure attention, not quality, maintenance or fitness for your use. Ages and "last push" figures are measured from the snapshot time, not from when you open the page.

Momentum

Each daily run stores every repository's star count. Gains per day, 7 days and 30 days compare today with the nearest stored day at least that far back. If snapshots are further apart than the period, the gain is scaled to the period, and if the nearest one is more than two weeks too old the figure is left out rather than guessed. A repository is "new to the list" for two weeks after it first appears. 19 snapshots are currently in use.

Weekly reports

Each weekly report compares the last snapshot of a week with the last one before that week began, and says which two dates those were. A finished week's report is stored and never recomputed, so its page stays the same. The current week's report is provisional until the week ends.

Maintenance figures

Latest release and open issue and pull request counts come from GitHub at snapshot time. They are shown as plain facts, not rolled into a score: a project with no releases may simply not use them, and a large issue count often means many users, not neglect.

Domains

Domains are assigned from an editorial taxonomy, not by GitHub. A repository scores points for matching topics, for whole-word matches in its name and description, and a little for its language. Language alone is never enough: a Python web framework is not filed under machine learning just for being Python. A repository with no clear signal lands in "Other" (831 today) instead of being forced into a category.

Similar repositories

Similarity is topic overlap, weighted so that rare topics count for much more than common ones. Two projects that both use raft are more alike than two that both use python. Repositories without topics have no similar list.

"Understand this repo" links

Every repository links to four independent sites that explain it, diagram it, turn it into a text digest for an AI assistant, or write a prompt that would rebuild it. They are third-party services that work by swapping github.com in the address for their own domain. This site only links to them and has no say over their availability or output.

Your lists, comparisons and stars

Saved lists and the comparison selection live in your browser's local storage. They are never uploaded. A share link carries the list inside the link itself, so whoever opens it gets their own copy.Check my stars reads a username's public stars straight from GitHub in your browser; nothing passes through this site.

Reuse

The full dataset behind this site is available as snapshot.json.