Reposby GlyphSources and method
Where every fact comes from, how new projects are found, and how stars per week are measured.
Which projects
The aim is every open-source data and AI project a knowledgeable reader would expect: every one with 10,000 or more GitHub stars that a curated list names (OSS Insight collections, the LF AI & Data and Apache project lists, the GitHub organisations of companies in the companies.fru.dev registry), every foundation-hosted data or AI project with 1,000 or more, every data or AI project a registry company leads with 5,000 or more, each registry company's own most-starred data or AI project from 1,000, and every relicensed data or AI project whatever its size. Lists, courses and books are left out.
Coverage today: 100% of the 10,000-star projects found so far are tracked; 0 found only by a GitHub topic search wait for review, and 0 named repositories are not read yet. Projects the run adds on its own are marked Unverified until a person checks their category and scope (104 now).
Growth, shown as "+N in the last 30 days", counts only projects, sources and license changes the daily run found after 2026-09-25. The curated seed and the hand-researched load that finished that day are the starting point, and later seed, research or backfill loads never count as growth.
Company coverage
Every company in the companies.fru.dev registry is enrolled and checked for a GitHub organisation with open-source data or AI projects, a few a day. 486 of 2,867 checked so far.
76%of companies in scope covered
154 covered, 48 with an organisation still to read. In scope: a company whose organisation is not read yet or has a data or AI project over the floor.
- Covered 154
- Organisation found 48
- Under the floor 143
- No organisation 140
- Out of scope 1
- Not checked yet 2,381
Bellwether companies
The companies a reader expects to find, checked first: 28 of 30 covered. A company counts from its most-starred data or AI project with 1,000 or more stars.
Where projects are found
549 sources so far, +282 in the last 30 days. The last run probed 8, added 16 and retired 0; it named 0 new repositories and added 1 projects. Also as JSON: /api/sources.
- GitHub topic llmGitHub topic59 added59 named
- Relicensing research (LICENSE commits and announcements)Research45 added45 named
- Apache projects (data, database and machine learning categories)Foundation list40 added362 named
- GitHub topic deep-learningGitHub topic25 added25 named
- GitHub topic machine-learningGitHub topic24 added24 named
- Artificial IntelligenceOSS Insight collection23 added27 named
- LF AI & Data landscape (hosted projects)Foundation list22 added56 named
- GitHub topic databaseGitHub topic16 added16 named
- Open Source DatabaseOSS Insight collection15 added15 named
- LLM ToolsOSS Insight collection12 added12 named
- Hugging Face on GitHubRegistry company on GitHub10 added19 named
- GitHub topic ai-agentsGitHub topic10 added10 named
- Time Series DatabaseOSS Insight collection9 added43 named
- Search EngineOSS Insight collection9 added12 named
- Nvidia on GitHubRegistry company on GitHub9 added9 named
- Stable Diffusion EcosystemOSS Insight collection9 added9 named
- AI Agent FrameworksOSS Insight collection9 added9 named
- GitHub topic large-language-modelsGitHub topic9 added9 named
- GitHub topic ragGitHub topic9 added9 named
- Message and StreamingOSS Insight collection8 added8 named
- ChatGPT AlternativesOSS Insight collection8 added8 named
- Coding AgentsOSS Insight collection8 added8 named
- Multimodal AIOSS Insight collection8 added8 named
- Business IntelligenceOSS Insight collection7 added15 named
- DeepSeek on GitHubRegistry company on GitHub6 added17 named
- Cloudflare on GitHubRegistry company on GitHub6 added15 named
- Graph DatabaseOSS Insight collection6 added14 named
- DeepMind on GitHubRegistry company on GitHub6 added11 named
- MLOps ToolsOSS Insight collection5 added5 named
- LLM DevToolsOSS Insight collection5 added5 named
- LLM Inference EnginesOSS Insight collection5 added5 named
- Meta on GitHubRegistry company on GitHub4 added30 named
- Microsoft on GitHubRegistry company on GitHub4 added25 named
- Vercel on GitHubRegistry company on GitHub4 added20 named
- Anthropic on GitHubRegistry company on GitHub, found by the run4 added17 named
- ByteDance on GitHubRegistry company on GitHub4 added10 named
- Apple on GitHubRegistry company on GitHub4 added9 named
- Distributed File StorageOSS Insight collection4 added4 named
- Go DatabaseOSS Insight collection4 added4 named
- Modern Data StackOSS Insight collection4 added4 named
Licenses and their history
The current license is read from the repository's LICENSE file. GitHub's own detection is shown too, and says NOASSERTION for custom or multi-license files (Elasticsearch, Redis, Terraform, MongoDB and others), so those licenses are read from the file itself. Every relicensing is dated by the commit that changed the LICENSE file, with a link to it, and linked to the company's announcement where one exists. Open source here means at least one of the licenses offered is approved by the Open Source Initiative; anything else is source-available. Not legal advice: the LICENSE file is the text that applies.
Forks, foundations and companies
Forks made after a license change (OpenSearch, Valkey, OpenTofu, OpenBao) are dated from their announcement or repository creation. Foundations come from the foundations' own lists: Apache projects from projects.apache.org, CNCF and LF AI & Data hosted projects and their maturity from their landscape files, and the PyTorch, OpenSearch, DuckDB and OpenSSF foundations from their own pages. The backing company is the one that leads or sells the project; companies carry their companies.fru.dev registry slug.
Stars per week
The daily run reads about a hundred repositories, the ones read longest ago, so each project is read about once a week, and keeps one snapshot of its stars, forks and open issues per week. Stars per week is the difference between two snapshots at least six days apart. Until the second snapshot, it comes from OSS Insight's weekly star history for the three weeks before tracking began, labelled as such on each project. OSS Insight counts from GitHub events and trails GitHub's own totals, so only its weekly differences are used, never its totals.
The daily run
Every day at 13:15 UTC. First discovery: OSS Insight's list of collections (a new data or AI collection becomes a source), GitHub topics that five or more tracked projects share, and the GitHub organisations of registry companies (the registry's own link, or an organisation named like the company whose website matches); a few sources are read each run and the repositories they name are read through GitHub search, then added when in scope. Every registry company not checked yet is enrolled as a candidate, and a few organisations of companies not covered yet are read each run, bellwether companies first. Then the projects read longest ago, up to eleven repositories per GitHub search call, the latest release for a rotating slice, and the foundation lists. A change in GitHub's license detection is logged and sent to the owner for review; the curated license moves only by hand, from a source. No model and no paid search are used.
License history found in the repository
A few projects a run, the run lists the commits that touched the LICENSE file and reads the file at each one. When two neighbouring commits read as different licenses and no dated relicensing already covers that date, it adds the change as Unverified, linked to the commit, until a person checks it.
History
Nothing that changes is overwritten silently. Releases, license detection, renames, archives, foundation listings and stars per week each write a row with the old value, the new one, when and the source; star counts go to weekly snapshots. The Changes page lists them and /api/changes serves them. /api/companies lists every backing company with its registry slug, domain and dated items for companies.fru.dev.
About this data
Projects from GitHub (repositories, releases and LICENSE commits), the foundations' own lists and company announcements. A daily run reads the hundred projects read longest ago (each about once a week), finds new projects in public lists and adds them marked Unverified until checked; license changes are set by hand from a source, or found in the LICENSE file's history and marked Unverified. The list grows as new projects appear. The LICENSE file is the text that applies. Not legal advice. Logos via logo.dev and GitHub; trademarks belong to their owners. Anonymous usage stats via Google Analytics. Use Suggest a correction on any page.