How trending works

Upstar ranks GitHub repositories by how fast they gain stars inside each topic, so a small project that is taking off in its niche is not buried under projects that already have tens of thousands of stars.

What is measured

For every tracked repository, Upstar counts the stars gained over three windows: the last 24 hours, the last 7 days and about the last 30 days. Star counts come from GitHub’s public data and are refreshed every hour; busy repositories are checked more often.

The score

Within a topic, repositories are ordered by stars gained ÷ √(stars before the period, at least 100). Growth counts relative to size: a project that goes from 50 to 800 stars this week (score 75) outranks one that goes from 6,600 to 8,400 (score about 22), even though the second gained more stars. Ties go to the larger gain.

A repository is ranked in a period only if it gained at least 20 stars in it, so a handful of stars on a brand-new repository cannot game the list.

Emerging repositories

The Emerging list and each topic’s Emerging tab look for acceleration rather than size: repositories that had fewer than 1,000 stars a week ago and are now gaining stars at least twice as fast as in a usual week. The usual week is the median of the 8 weeks before; a usual week under 10 stars counts as 10, so going from 0 to 25 stars is not enough on its own. A repository must be at least 30 days old to have a usual week to compare with, and gain at least 20 stars this week. The list is ordered by that ratio, and goes through the same review and content rules as the rankings.

Topics

Each repository is placed in topics from its GitHub topics, its main language and words in its description. Upstar tracks a fixed set of topics (Emulation, Self-hosted, Rust, MCP…); a repository can rank in several. Forks are not ranked.

Finding new repositories

Every hour Upstar reads a sample of GitHub’s public event stream to spot repositories that start gaining stars, and every few days it searches each topic for recent projects. A repository usually appears within a day or two of taking off.

Keeping the lists honest

• Star farming. Bought or automated stars leave traces: few forks, issues, pull requests or contributors for the number of stars, few watchers, very young owner accounts, stars that GitHub later removes. Repositories that look farmed are left out of the rankings.
• Human review. Doubtful cases are not removed automatically: they stay in topic rankings but wait for a person to review them before being featured on the home page, in digests, alerts or social posts.
• Adult content is not ranked. Repositories about adult content are detected from their description, topics and README and left out; tools that detect or filter such content are not affected.
• Lures such as “cracked” software or fake license guides are sent to review or blocked.
• Anyone can flag a repository with the Report this repo link on its page.

Health

Repository pages also show health signals: commit regularity, maintainers, responsiveness to issues, releases and project hygiene. They help tell a maintained project from a burst of attention. How health is computed.

What Upstar is not

Stars measure attention, not quality. Upstar is not affiliated with GitHub, Inc. and does not endorse the projects it lists.