Loading
About the data
Published. Read this before quoting any figure.
Every country figure is an observed minimum, not a count. Worldwide, the source dataset places roughly 40% of active developers in any country. That is a rate for the whole dataset, not for Nigeria or any other country: how many of the unplaced developers live in a given country cannot be observed, so we do not estimate it, and the true number is higher by an amount we cannot state.
This is not a gap that more data closes. The unattributed developers have not declared a location anywhere public. There is no dataset that knows where someone lives when they have not said. Asking them — the confirmation campaign — is the only lever, and its ceiling is the response rate.
59% of "active" Nigerian developers in 2025 contributed fewer than ten days across the whole year. There is a structural cause: 419 repositories with 20 or more contributors and fewer than 50 stars, 101 of them with 50 or more contributors. Normal open-source projects do not look like that. It is the signature of paid bounty and hackathon programmes, and the ecosystems running large Africa contribution programmes sit high in the country tables.
We treat this as a finding, not a defect — paid contribution programmes are a real on-ramp, and that is more interesting than a headcount. But it is why every headline figure is segmented by contribution intensity, and why an unsegmented number should not be quoted from this site.
Electric Capital's repository taxonomy is community-submitted. African projects that have not been submitted are invisible to it. This pushes figures down, not up.
Documentation, translation, design, review and community work are not counted. This undercounts exactly the kind of contribution through which many people begin, and undercounts it unevenly across communities.
The declared tier rests on a profile string. People write cities, joke entries, and nothing at all. Geocoding quality is good — "Lagos" resolves to Nigeria rather than Portugal in the overwhelming majority of cases — but a declared location is a claim, not a verified fact. Locations read from GitHub profiles are today's, applied to every quarter the developer was active: someone who moved is counted where they say they live now, as with Electric Capital's own locations.
City cells below a disclosure threshold are rolled into an "other" bucket, because a count of one in a named city in a named country, combined with public GitHub data, identifies a person. City breakdowns are therefore incomplete by design.
Upstream rebuilds daily and recent periods are the least settled. Figures for the most recent quarters are the most likely to move between releases. Where the data is not yet trustworthy the period is withheld rather than published provisionally.
A developer counts for an ecosystem when they contribute to any project the source files under it, including its sub-ecosystems. A single popular project filed under several chains can move several ecosystems' shares at once: in September 2026 one AI agent repository was tagged to nine chains, and 16 developers in the tracked markets contributed to it. Shares are therefore a measure of contribution to what the source files under each ecosystem, not of what developers would call their own focus.
The source keeps adding older projects, and each addition brings in past contributions: 90% of the developers whose first contribution fell in 2019 were only added later. So the latest months' new developers and repositories are undercounted and rise in later releases. They are drawn as provisional. The "newly visible" series, which separates developers who appear late from genuine newcomers, is an approximation built from the date each project was added.
The current year's active developers are counted to date and are never compared with a whole earlier year.
Our own 2025 Nigeria report quoted "4% of global Web3 developers". That figure is not reproducible from Electric Capital's open data, which gives 0.77% on the closest equivalent definition, and their own published figure is 3.3%.
The mechanism is traceable: roughly half the gap is which repositories count as crypto, and the rest is an eligibility filter based on market-capitalisation data that is not in the open dataset. We cannot replicate it, so we do not publish shares of the world. Absolute counts and growth rates are unaffected.
A figure published here will therefore differ from the 2025 report. That is a methodology change, not a collapse in developer numbers.
An early internal memo computed "sustained" as 10 or more contribution days, which includes heavy contributors. Figures on this site use a true partition (under 10, 10 to 49, 50 or more). A sustained figure quoted from that memo will be higher than the one here, by exactly the heavy count, and the two should not be compared.
v1 measures Web3 only. No published estimate of African AI open-source contributors exists anywhere, which is why that measurement is being built rather than sourced. Until it ships, this site says nothing about AI.
They cover only the hackathons collected (a reviewed list on Devfolio, later Colosseum, and whatever organisers send), only participants on submitted projects, and only those whose public GitHub profile names a place in one of the five countries. None of the Devfolio events collected was held in Africa; African participants there joined online or travelled. Organiser reports are self-reported and cannot be deduplicated across events. Locations are read today, so someone who has moved is counted where they live now.