How the guide works

Methodology

Public ecosystem signals help us notice change. Sourced editorial judgment determines what is useful enough to publish.

Trending is measured momentum

We store aggregate GitHub observations and compare eligible snapshots over approximately 7, 30, and 49 days. The trending page ranks catalog projects by net stars gained in the selected period, not lifetime totals. We show the actual observation dates: the seven-week baseline must be within three days of 49 days. This is not an index of all GitHub repositories or a count of every individual starring event. New projects are labeled as early signals; incomplete windows are never extrapolated. Other discovery surfaces may use our separate, category-normalized momentum score, which also considers relative growth and recent repository activity.

Recommended is editorial

A recommendation is a separate human decision. It reflects usefulness, technical approach, developer experience, maintenance, and the quality of the tradeoff—not merely popularity. A project can be trending, recommended, both, or neither.

Guardrails

Stale, private, archived, anomalous, or manually excluded repositories cannot automatically take a ranking slot. We publish the evidence window and confidence instead of presenting an unexplained score. The precise weighting remains internal to reduce gaming.

Pricing and open source

Open source describes licensing, not the price of a hosted service. Pricing labels summarize a meaningful early production use case and link to the source checked by the editorial team. “Unknown” is preferable to a guess.

Stacks and corrections

Stacks are authored around a target user and responsibility map. Each choice needs a rationale, cost and ownership context, alternatives where they materially differ, explicit tradeoffs, and current sources. If a fact is wrong or stale, use the corrections page.