Snapshot 2026-08-02
How these numbers are made
Every leaderboard in this space presents its ranking as fact. None of them tell you what the ranking cannot see. This page does both.
What is counted
A developer’s score is their public plus private contributions over the trailing twelve months. Private contributions are the count GitHub itself exposes — a number, never the underlying work. Ranking uses contributions first, followers second, and login alphabetically as the final tie-break, so equal scores never swap places between snapshots.
How developers are found
Candidates come from GH Archive, the public record of every public GitHub event. Streaming a rolling seven-day window finds everyone who did something visible, costs no API budget at all, and takes seconds rather than the hours a search-based sweep needs. Only the filtered survivors are then looked up through GitHub’s API.
Every board — worldwide, country and city — is then derived from that single hydrated set by grouping it on the parsed location. That is what removed the old ceiling: a location-scoped search returns at most a thousand results per query, so the previous pipeline could never rank more than a thousand people anywhere, however long it ran.
The limit worth knowing: the archive only records public activity, so somebody working almost entirely in private repositories does not appear in it. A smaller search pass covers that gap by contributing names for the next hydration to fetch. Accounts firing thousands of events at a single repository are dropped as automation before anything is looked up.
Where location comes from
GitHub has no structured country or city field — only a free-text line that people fill in however they like. Both country and city are parsed from that one string, and a city is kept only where at least eight tracked developers agree on a name. “Earth”, “remote” and “/dev/null” are discarded rather than guessed at. This is the single largest source of error on the site, and it affects every competitor equally.
How deep the rankings go
The board is the whole snapshot, not a top hundred: every ranked developer is reachable, sortable by contributions, followers or streak, and searchable by login, name, company or location. The ceiling is 250,000 developers, and it is a storage decision rather than an API one — snapshots are committed to git so each one can be diffed against the day before, and a corpus much past that size would be rewritten wholesale on every crawl.
Measured against estimated
Contribution totals and follower counts are measured, for everyone. The day-by-day calendar is not: it is by far the most expensive thing to fetch, so it is bought for the top of the board and estimated below that. An estimated calendar is derived deterministically from the measured total — the same login always produces the same calendar, and the total always matches exactly.
This matters most for streaks, which can only be read off a calendar. Ranking by streak therefore ranks measured and estimated values together, and every estimated row is marked with ≈— in the text, not only in the colour. An estimated streak is a plausible shape, not a fact about that person’s week. The scheduled crawler replaces estimates with real calendars as it reaches them.
Organizations
The organization board ranks employers by the combined contributions of tracked developers who name them in their profile, not by follower count. It is drawn from a free-text company field, so it reflects what people write, including spelling variants.
Accounts we exclude, and why we show you anyway
7 accounts exceeded 300,000 contributions in twelve months — over 820 every day without a break. That is not a person; it is a bot committing under a personal account. They are excluded from every ranking. Most leaderboards either let these accounts sit at number one or drop them silently. Here they are, so you can judge the rule for yourself.
- @0x3EF818.9M
- @u1i1.7M
- @shahinpr951.9K
- @CodeWithEugene921.5K
- @haythemgalelem586.9K
- @bugbounted570.5K
- @farhan7reza7505.7K
Where this snapshot came from
Bootstrap snapshot mapped from the public gayanvoice/top-github-users dataset. Replaced by this repository's own crawler on its first scheduled run.
Bootstrap data is mapped from gayanvoice/top-github-users, a public dataset, with thanks. This repository’s own crawler replaces it on its first scheduled run.
Reproducing it
The crawler, the schema and this site are all in one repository. Snapshots are committed as JSON, so any ranking on this site can be diffed against the day before it.
Design
Motion patterns come from rejouice-patterns, scaled down deliberately: this is a data site, so the display type is roughly a third of the corpus default and density carries the page. Every animation is disabled under prefers-reduced-motion, and every route is readable with JavaScript switched off.