The repositories
building what’s next.

Version 2: 35 points relevance, 20 adoption, 20 maintenance, 15 momentum and up to 10 community points. Category tabs rank up to 100 eligible projects independently; the overall list shows the highest 100 scores across categories. Search and time filters do not renumber ranks.

Adoption compares log-scaled stars (50%) and forks (25%) against eligible peers in the same category, plus contributor identities (25%, capped at 5,000). Relevance uses the displayed classification reason; it can contain judgment or automated classification errors.

Maintenance splits equally between recent push activity, recent stable releases, sampled active contributors and sampled maintainer response coverage. Pushes within 30/90/180/365 days score 100/75/50/25%; older pushes score zero. Stable releases within 90/180/365 days score 100/75/50%; older sampled releases score 25%. Active contributor credit grows logarithmically up to 20 accounts. Maintenance checks rotate through category leaders; evidence older than 42 days is neutral. No sampled release or eligible issues is neutral, not a failure.

Momentum uses actual observations 6–35 days apart, preferring a roughly monthly baseline. Net star and fork growth is scaled to 28 days with baseline smoothing, then capped. New projects wait for history; missing signals use 50% of their allocation. Scores can move as evidence becomes available.

One signed-in vote per account per repository across all views. Vote points grow logarithmically and cap at 100 votes / 10 points. Inactivity of 181–365 days reduces the total by 10%; longer inactivity, archival or a maintenance notice reduces it by 25%. Missing evidence and limited samples are not proof of quality. This selected search pool does not cover all GitHub projects.