In the past two weeks, 18 news organizations created or opened 19 public repositories on GitHub. Highlights below.
Highlights
Local Angle announced Backfield on Tuesday. It is an open-source platform that turns news stories into structured data: geocoded locations, people and their quotes, and a knowledge graph of the people, places and organizations in a newsroom’s coverage. Along with the platform, Local Angle released a live demo, a cookbook of tutorial notebooks and a self-hosting guide.
I asked Chase Davis why he decided to open source the project. He cited a desire to contribute back to the news-nerd community: “It exists largely because of the support and belief of news leaders who care deeply about improving and sustaining our industry,” he said, adding, “Backfield is for builders, and builders are my people.” Read the full interview.
ProPublica released SSI Skills, agent skills for what builder Aaron Brezel calls “spreadsheet inference”: running the same AI prompt once per row across hundreds of documents, records or links. The first skill turns a reporting question into a prompt, along with a written record of the definitions and edge cases behind it. It installs in Claude Code or the Claude apps, and a hand-ported Gem covers Gemini. It is a companion to ProPublica’s SSI Toolkit add-on for Google Sheets released earlier this year — and which Brezel has been actively maintaining. The repo ships an AGENTS.md.
The Hechinger Report rebuilt its College Financial Health Tracker and opened the repository that serves it. The tracker follows closures, layoffs, program cuts, accreditation actions and research-funding losses at 1,890 four-year colleges. It launched with the story “More than half of colleges show signs of money problems. Students will suffer the consequences.” One interesting thing I noticed: The GitHub repo includes multiple licenses. Its data is free to reuse under a Creative Commons “CC BY 4.0” license. The pipeline, brand assets, and front-end code are not licensed for re-use and published “so readers can inspect how the College Financial Health Tracker is built.”
Fiquem Sabendo, the Brazilian public-records outlet, published a searchable tool on federal tax breaks by company, 2015 to 2024, along with the pipeline behind it. It loads a decade of tax-expenditure microdata from Brazil’s Transparency Portal into DuckDB and lets readers rank and search 68,190 corporate groups by name or company registry number. The build also writes a copy of the page for use as a Claude artifact.
City Bureau published a playbook for building and repairing City Scrapers with coding agents, written for Code the Dream engineers in a pilot. It includes three skills, for diagnosing a scraper alert, recording the diagnosis and taking a repair through to a pull request.
The Philadelphia Inquirer opened a self-hosted image search engine for news photo archives. It fetches images from public URLs, has a multimodal model describe each one, and stores the embeddings in Chroma so editors can search the archive by description. It has a Gradio interface, a REST API and an MCP endpoint, and works with Azure, OpenAI, Anthropic or Gemini models.
Diagram Chasing published two more Indian government datasets as open files: readings from 1,747 Central Water Commission river gauges and dams, with daily data back to 1900 (sparse before 1975), and weather balloon launches from 57 India Meteorological Department stations since 2009. The data powers dashboards for flood forecasts and balloon launches. The flood-gauge README notes that its code and documentation were written with AI assistance.
Montana Free Press opened its analysis of PAC money behind “Who’s drawing the PAC power in Montana’s 2026 congressional races?” It pulls traditional-PAC donations to Montana’s congressional candidates, and donations made by the candidates’ own leadership PACs, and classifies each donor against a snapshot of the FEC’s committee file.
iMEdD mapped how far every part of Greece is from the nearest secondary school, measured along roads, for “How close is the nearest school? Mapping distances across Greece.” The repo includes the isodistance zones as GeoJSON, a layer of islands with no school, and a notebook to rerun the calculation for one area.
The Markup published the data behind their story “Defense tech has unleashed a new weapons boom in Southern California.” The data covers every Defense Department contract obligation to a California recipient from fiscal 2002 through 2026, pulled from USASpending.gov, with each winner’s congressional district and county. They also published a Marimo notebook that fact-checks the story’s key findings.
Deutsche Welle published the data and code behind “What really helps Europe deal with drought,” Kira Schacht’s analysis of where European countries could save the most water. It combines Eurostat and European Environment Agency figures on water use by sector for 37 countries.
Correctiv opened the code and data behind “Discover How Drought is Affecting your Region,” its interactive drought atlas of Europe from June.
Inside Climate News opened the R analysis behind its Sept. 1 investigation of EPA enforcement under the second Trump administration. It draws on the EPA’s enforcement database and on Office of Personnel Management data on agency staffing.
By the Numbers
Beyond new repos, 87 news organizations made a combined 5,423 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 | 1,290 |
| Bloomberg | 384 |
| Freedom of the Press Foundation | 331 |
| Buried Signals | 298 |
| Vox Media | 240 |
| ICIJ | 227 |
| OpenSanctions | 189 |
| OpenElections | 146 |
| Bad Idea Factory | 134 |
| Big Local News | 127 |
This roundup is assembled with the help of open-journalism-bot, which monitors public GitHub activity from ~420 news organizations and posts new repositories to @openjournalism.news on BlueSky. Commit counts shown here exclude commits we identified as automated.