In the past two weeks, a dozen news organizations created or opened public repositories on GitHub. Highlights below.
Highlights
Deutsche Welle published the data, code, and methodology behind its analysis of pesticide sales, use, and impact across the EU from 2014 to 2024. The Jupyter notebooks combine figures from the UN Food and Agriculture Organization, Eurostat, the European Environment Agency, and Pesticide Action Network International to track sales against the EU’s proposed (and later shelved) reduction targets. The analysis, a European Data Journalism Network collaboration, found pesticide sales fell in 20 member states since the 2015–2017 baseline, with Italy down 33%. Research by Ana Muñoz Padrós, who also wrote the DW Story, “EU is failing to cut pesticide use”.
The project’s README is especially good, and lays out the research findings, including visualizations. Well worth a read.
Montana Free Press released the data and analysis behind its reporting on spending in Montana’s contested Republican legislative primaries. The repo holds committee contribution and expenditure records going back to 2018 alongside Python scripts that trace cash-flows by committee. The repo backs a May 29 story finding that four PACs accounted for 96% of the spending across the state’s 43 contested GOP primaries.
The Trace published dataset-butler, a small GitHub Actions scheduler it built to keep its public datasets current. Redivis, the data platform the newsroom uses, has no built-in way to run a notebook on a schedule, so The Trace wires up a cron-triggered Action that calls the Redivis API and refreshes a dataset on a set cadence. Two instances opened during this window — one updating a tracker of immigration-related shootings hourly, another refreshing a University of Virginia gun-violence dashboard weekly.
The Philadelphia Inquirer released Dewey, a Model Context Protocol server that exposes its news archive to AI assistants and other tools. It offers a search_archive function backed by Azure AI Search that returns article chunks filtered by date, author, and other criteria. The repo ships an AGENTS.md.
Haaretz published the first stage of an AI editing service that flags Hebrew language errors in HTML. The FastAPI app takes an HTML document and returns it with word-level corrections wrapped in <mark> tags, routing inference to Claude, GPT, or Gemini through LiteLLM so no model weights ship in the image. Haaretz describes it as the first stage of a broader “editors-agent” project.
Fiquem Sabendo, a Brazilian freedom-of-information group, published an archived static copy of WikiLAI, a 232-page MediaWiki knowledge base on access-to-information law and transparency resources. The archive preserves the original URL structure so external links keep working, and ships setup instructions for others preserving MediaWiki sites the same way. Most of its commits are credited to Claude, and it carries a committed .specstory/ directory.
AJ Digital opened a terminal dashboard that wrangles React Native’s development workflow. Built in Rust, the tool pulls git worktree management, the Metro bundler, and device deployment into a single terminal window. The repo ships a CLAUDE.md, and about a fifth of its recent commits are credited to Claude.
By the Numbers
Beyond new repos, 81 news organizations made a combined 3,458 public commits to GitHub during this period. The most active by commit count (excluding, as best we can, commits done by bots, gh-actions, or cron):
| Organization | Commits |
|---|---|
| The Guardian | 850 |
| Freedom of the Press Foundation | 344 |
| ICIJ | 166 |
| OpenNews | 133 |
| OpenAleph | 99 |
| OpenStates | 94 |
| OpenSanctions | 89 |
| MuckRock | 87 |
| The Pudding | 87 |
| Spotlight PA | 86 |
This roundup is assembled with the help of open-journalism-bot, which monitors public GitHub activity from ~360 news organizations and posts new repositories to @openjournalism.news on BlueSky. Commit counts shown here exclude commits we identified as automated.