In the second half of March, 20 news organizations created or opened 26 public repositories on GitHub.
Highlights
ProPublica released gas-ssi-toolkit, the source code for their SSI Toolkit, a Google Sheets add-on for AI-assisted investigations. It includes curated AI “recipes” — reusable workflows that run over spreadsheet rows — plus configurable Gemini API inference with support for text, multimodal files, and tool calling. It can also bulk-import files from Google Drive, extract text from Docs, PDFs, and images via OCR, and do reproducible dataset sampling for rapid prompt refinement.
The Data Liberation Project and The Marshall Project are collaborating on banned-books, an effort to collect and standardize lists of books banned in state prison systems across the country. The repo builds on The Marshall Project’s 2023 investigative series on prison book censorship, which included a reporting toolkit used by dozens of local newsrooms. They’ve obtained data via public records requests from 12 states so far, and they’re seeking volunteers to help clean and standardize the data. (Disclosure: I’m on the board at Muckrock, which runs the Data Liberation Project).
Inside Climate News published the data and R code behind their analysis of changes in the federal workforce during the first year of the second Trump administration. The analysis uses monthly employment snapshots from the Office of Personnel Management to track hiring and departures at federal agencies, and supported two stories on staffing cuts at the EPA and Department of the Interior.
Deutsche Welle published the data and methodology behind their investigation into whether an increase in female representation in African parliaments has improved gender equality. The analysis, by Ana Muñoz Padrós from DW’s data journalism team, uses UN and World Bank data to test correlations across education, financial independence, and bodily autonomy indicators.
CalMatters released a pair of tools for extracting and analyzing California political disclosures: fppc700-download, a CLI that bulk-downloads Form 700 financial disclosure PDFs from the California Fair Political Practices Commission database, and fppc700-extract, which uses OCR to pull structured JSON from those PDFs. Together they automate what has historically been a manual process of reviewing state officials’ financial interests, income sources, and gift disclosures. The extract tool works both as a standalone CLI and as a Python library, and targets documents from 2024 onward when electronic filing became mandatory.
MuckRock released foia-coach, a prototype LLM-powered tool for helping people with public records requests. It uses a RAG (retrieval-augmented generation) system with per-state jurisdiction knowledge, supporting multiple LLM providers including OpenAI and Gemini. The tool includes a Django API backend and SvelteKit frontend.
The Philadelphia Inquirer released pmn-ai-workflow, a CLI tool that automates their engineering team’s development workflow from Jira ticket to pull request. Point it at a ticket and Claude Code reads the ticket details, pulls related Figma designs and Confluence documentation, creates a branch, implements the feature, runs lint, opens a PR, and posts results back to Jira. If the AI encounters unclear requirements, it posts questions directly to the Jira ticket as comments.
Local Angle released agate-ai-demo, a public demo of their Agate tool, which uses large language models to turn news articles into “structured, durable knowledge.” The demo packages a complete stack — UI, API, worker, PostgreSQL, and Redis — so newsrooms can explore structured-journalism workflows locally. Initial work began under the Lenfest AI Collaborative and Fellowship Program at the Minnesota Star Tribune and has since continued in collaboration with Chicago Public Media and the Reynolds Journalism Institute.
Montana Free Press released two repos: candian-border-crossing-changes-24-25, a Python analysis of Bureau of Transportation Statistics data tracking changes in U.S.-Canada border crossing activity at Montana ports; and olympian-birthplace, a data analysis mapping which U.S. states have produced the most Olympic athletes.
The Pudding published the code behind A Journey Through Infertility, an interactive scrollytelling piece by Lam Thuy Vo about IVF told from two parallel perspectives — the mother and the baby. Built in SvelteKit with D3, it uses an isometric illustrated world (inspired by the game Monument Valley) that zooms and pans through SVG scenes as the reader scrolls, with sprite-based force-layout visualizations and a cost comparison table showing IVF expenses across seven U.S. cities.
Recovered Factory released missouri-vsr-tool, a bilingual (English/Spanish) SvelteKit web app for visualizing Missouri’s Vehicle Stops Report data — police traffic stop statistics broken down by agency and year. Users can search for specific agencies, explore metric breakdowns, compare neighboring jurisdictions, and download datasets.
Buried Signals released skills, a Claude Code plugin marketplace for journalists and investigators. It includes three toolkits: Spotlight, an OSINT investigation orchestrator that guides verified investigations from lead to findings and archives them into an Obsidian vault; OSINT, a catalog of 150+ investigation tools from Bellingcat, IDI, and DigitalDigging covering geolocation, social media analysis, satellite imagery, and more; and Storytelling, a framework for building scroll-driven narrative experiences with SvelteKit.
ABC News Australia released service-petrol-prices, a service that fetches and stores gas price data for tracking fuel costs over time.
USAFacts released poc-data-ai-ecosystem, a proof-of-concept manifest-driven architecture (MDA) for government data ingestion, processing, and visualization. The system includes a data pipeline with manifest-driven configurations, a FastAPI backend, and a React frontend with dashboards for browsing agencies, assets, and workflows.
By the Numbers
Beyond new repos, 78 news organizations made a combined 6,810 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 | 1,210 |
| Bellingcat | 407 |
| Freedom of the Press Foundation | 293 |
| Minneapolis Star Tribune | 271 |
| The Economist | 254 |
| OpenSanctions | 122 |
| PRX | 92 |
| City Bureau | 79 |
| OpenNews | 77 |
| ABC News (Australia) | 65 |
Data comes from the Open Journalism Bot, which monitors ~360 news organization GitHub accounts. Follow @openjournalism.bsky.social for real-time alerts.