More than a dozen news organizations opened public repositories on GitHub during the past fortnight. Highlights below.
Highlights
Three newsrooms, Lighthouse Reports, Wired, and The Independent, published the data and analysis behind Asylum by Algorithm, a joint investigation into the facial age estimation software the U.K. Home Office uses to judge whether asylum seekers are adults or children. Working from NIST’s published evaluation of Cognitec’s algorithm, Maddy Varner from Wired, Gabriel Geiger from Lighthouse Reports, and their colleagues parsed the report-card data to measure two error rates: how often 16-year-olds are wrongly flagged as adults, and how often 25-year-olds are read as minors, broken down by sex and geographic region. Lighthouse Reports also published a detailed methodology.
The Pudding shipped two interactives that had been in development for a while. “A History of Menus is a Menu of History,” by Matt Daniels, tells the story of American restaurant dining in ten dishes, drawing on the New York Public Library’s Buttolph collection of historic menus. They made the code and data behind the project public. They also opened up the source code for “The Lawn Mowing Experiment,” by Russell Samora, which turns mowing a lawn into a game about finding the most efficient route — a playable take on the Traveling Salesman Problem. The project is a running experiment with results expected soon.
Inside Climate News published the code and data behind its analysis showing that Democrats are embracing “climate hushing”. Peter Aldhous searched an open archive of congressional press releases for mentions of climate change, energy affordability and clean energy. The archive, Congress Press, is an open archive of nearly half a million congressional press releases from 2016 to 2026 built by Derek Willis. An open newsroom repo built using open data from a journalist!
The Guardian published the data behind its investigation into the fees driving up American apartment rents. The dataset breaks down the more than one million U.S. apartment units overseen by Greystar — the country’s largest apartment owner and manager — by city and by state as of December 2025, drawn from a Private Equity Stakeholder Project analysis of Yardi Matrix data.
The Washington Post released the code behind its analysis of whether AI chatbots are politically biased. Built by Kevin Schaul, the tool feeds politically charged questions drawn from ModelSlant to models like ChatGPT and Gemini, then scores whether each response offers left-leaning arguments, right-leaning arguments, or both.
The South Florida Sun-Sentinel published an auto-updating map of active Florida wildfires and released the code behind it. A scheduled GitHub Action runs every half hour, pulling active-incident and smoke data from the National Interagency Fire Center into a GeoJSON feed that drives a Leaflet map.
OpenSanctions released Kolkhoz, an early-stage pipeline by Johan Schuijt that, in its own words, “turns the internet into lists of politicians.” It pairs with Pravda, the web-archiving service OpenSanctions released a few weeks ago. Kolkhoz sends URLs to Pravda for snapshotting, feeds the captures to an LLM to extract structured person-and-position pairs, and exports the results as a FollowTheMoney entity stream. The repo ships an AGENTS.md.
Recovered Factory, the shop of data journalist David Eads, opened a public-interest tool tracking the 287(g) agreements that let local police and sheriffs carry out federal immigration enforcement for ICE. Eads also published a pipeline and bilingual map viewer for Cali, Colombia’s municipal geodata, which downloads all 357 layers from the city’s GeoServer and converts them to web-friendly formats. The bulk of the commits in both projects are credited to Claude Code.
City Bureau released a preview-and-validation tool for civic event feeds produced by its City Scrapers project, which collects public-meeting data for civic-tech volunteers. Built by Pat Sier and Rajiv Sinclair, the React app renders scraper output as a searchable list and calendar and flags missing or malformed fields before the data is published. It ships both a CLAUDE.md and an AGENTS.md, and a share of its commits are credited to Claude.
Al Jazeera published Mission Control, a local dashboard by Victor Hugo dos Santos that reads the session transcripts Claude Code and Codex leave on disk and reports per-project and per-feature token and cost analytics. It correlates each coding session to a feature via the project’s Git branch name, stores everything in a local SQLite database, and binds only to localhost. The tool ships a CLAUDE.md.
By the Numbers
Beyond new repos, 88 news organizations made a combined 4,341 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 | 1,023 |
| Freedom of the Press Foundation | 396 |
| ICIJ | 392 |
| MuckRock | 180 |
| OpenSanctions | 148 |
| CBC Radio-Canada | 140 |
| PRX | 101 |
| Recovered Factory | 99 |
| Los Angeles Times Data Desk | 92 |
| The Pudding | 88 |
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.