data study · 2026
The State of Open Source, 2026
We analysed every project in this registry — 4,120 open source codebases with 40.8 million combined GitHub stars — to answer four questions: who gets the attention, what licences projects actually choose, which languages carry the ecosystem, and how much of it is still alive.
1. Attention is concentrated — but less than you would guess
The top ten projects by stars command 6.5% of all stars in the registry. The top hundred hold 27.9%. That is real concentration, but it is not a winner-takes-everything market: the remaining 72.1% of stars are spread across more than 1,700 projects, and the median project sits at 3,085 stars — comfortably above the threshold people use to judge whether a project is "real".
| star tier | projects | share |
|---|---|---|
| 50,000+ | 148 | 3.6% |
| 10,000 – 49,999 | 764 | 18.5% |
| 1,000 – 9,999 | 2,389 | 58.0% |
| under 1,000 | 819 | 19.9% |
2. Permissive licensing won — but copyleft is not dead
MIT and Apache-2.0 together cover the majority of licensed projects, and permissive terms dominate outright. Yet GPL-family licences (GPL-3.0, GPL-2.0, AGPL-3.0, LGPL) still account for a substantial bloc — and AGPL-3.0 in particular shows up heavily among self-hosted products, where the licence is doing commercial work rather than ideological work.
| licence | projects | share | avg stars |
|---|---|---|---|
| MIT | 1,690 | 41.0% | 10,023 |
| Apache-2.0 | 1,095 | 26.6% | 9,665 |
| AGPL-3.0 | 479 | 11.6% | 8,849 |
| GPL-3.0 | 338 | 8.2% | 7,274 |
| BSD-3-Clause | 88 | 2.1% | 8,142 |
| GPL-2.0 | 70 | 1.7% | 7,043 |
| MPL-2.0 | 58 | 1.4% | 11,890 |
| BSD-2-Clause | 31 | 0.8% | 12,062 |
| LGPL-3.0 | 27 | 0.7% | 2,862 |
| LGPL-2.1 | 14 | 0.3% | 4,402 |
3. TypeScript leads on count, C++ leads on stars
TypeScript is the most common primary language in the registry, ahead of Python and Go — a reflection of how much of the modern tooling layer is web-adjacent. But by average stars, compiled and systems-level languages pull ahead: C++ projects average more stars per repository than any other language here, with Python close behind. Popularity by count and popularity by attention are measurably different things.
| language | projects | avg stars | median |
|---|---|---|---|
| TypeScript | 998 | 11,716 | 3,030 |
| Python | 698 | 11,589 | 3,558 |
| Go | 575 | 9,612 | 3,648 |
| JavaScript | 339 | 11,026 | 2,996 |
| Rust | 279 | 10,657 | 3,922 |
| Java | 197 | 8,016 | 3,334 |
| PHP | 161 | 5,125 | 2,224 |
| C++ | 127 | 11,056 | 3,561 |
| C# | 118 | 5,074 | 2,330 |
| C | 71 | 9,381 | 3,571 |
4. AI is the smallest category by count and the biggest by attention
Ranked by number of projects, AI and machine learning is mid-table. Ranked by average stars, it leads every other category in this registry by a wide margin — roughly 3.6× the category with the lowest average. Whatever the long-term shape of the ecosystem, the current attention distribution is unambiguous.
| category | projects | avg stars | total stars |
|---|---|---|---|
| AI & Machine Learning | 448 | 16,694 | 7.5M |
| Developer Tools | 569 | 13,666 | 7.8M |
| Productivity & Utilities | 483 | 9,930 | 4.8M |
| Infrastructure & Operations | 812 | 9,185 | 7.5M |
| Miscellaneous | 442 | 8,831 | 3.9M |
| Data & Analytics | 330 | 8,049 | 2.7M |
| Content & Publishing | 302 | 7,734 | 2.3M |
| Security & Privacy | 201 | 6,954 | 1.4M |
| Business Software | 464 | 5,673 | 2.6M |
| Community & Social | 69 | 4,576 | 0.3M |
5. The ecosystem is alive
78.5% of tracked projects received a commit in the last 30 days, and 1,859 were created within the last three years. The oldest repository in the registry dates to 2008. The stereotype of abandoned open source does not survive contact with the data at this star threshold — though it is worth saying plainly that our inclusion threshold (1,000+ stars) selects for projects that already found an audience.
Methodology
- Sample: 4,120 projects tracked in the Opensourcedex registry on 2026-09-15
- Source: GitHub REST API — stars, forks, primary language, licence, creation and last-push dates
- Inclusion: 1,800 projects pass an automated gate (≥500 stars, a commit in the last 12 months, OSI-approved or fair-core licence, product signal); 97 were added by hand or via the submission queue
- Star counts are point-in-time; they move daily
- Licence is GitHub's classifier;
NOASSERTIONis excluded from the licence table because it means "no standard licence detected", not a licence choice
The dataset is queryable by agents through our public API:
GET /api/search?q= · raw aggregates available on request.