Over the past fortnight, 13 news organizations published 15 new open-source repositories — investigative data releases, election-service pages, and a newsroom-built Slack agent. Highlights below.
Highlights
I was excited to see KQED’s prototype boilerplate, a Vite/React starter written explicitly for AI-assisted prototyping by non-developers — it ships a CLAUDE.md and Cursor rules, and its README assumes the reader is pairing with a coding agent. “Many people using this repo are non-technical and are learning to build with AI assistance,” it says to Claude, “favor clarity and small, reviewable steps over cleverness.”
Deutsche Welle released the data and analysis behind “Strait of Hormuz traffic: Who is taking the risk?”, Farah Bahgat and Gianna-Carina Grün’s investigation of which ships kept moving through the strait during the conflict-driven collapse in traffic. The team measured vessel presence in Global Fishing Watch’s AIS data, then looked up each ship’s registered owner and manager in Equasis and MarineTraffic to see who was still taking the passage.
Flatwater Free Press published the calculator it built to help readers determine if they can “afford an ‘affordable’ home in Nebraska”, part of its July 31 story “Priced out of The Good Life.” Ben Vankat’s tool takes your income and county and applies “a common rule of thumb” — a home is affordable at roughly 2.5 times household income or less — against county median home values from the American Community Survey, showing how far out of reach the median home has drifted in much of the state.
Spain’s Civio built pi-dad, a minimal Slack agent harness that forwards @-mentions and DMs to an LLM and posts the reply — designed for local, privacy-respecting model servers like LM Studio, though it works with cloud providers too. The agent can run shell commands in a sandbox, keeps per-thread conversations and Markdown memory files, and discovers reusable “skills” from its workspace. It’s a successor-in-spirit to the retired pi-mom harness, and its recent commits are all credited to Claude.
Bloomberg open-sourced Causal-TS, a handy Python framework for discovering causal relationships in time-series data. It bundles four algorithms suited to different data conditions, including nonstationary data and large variable sets — the kind of high-dimensional data common in financial and economic analysis. They published full documentation in ReadTheDocs.
OpenSanctions open-sourced a tool in “early R&D” called Kolkhoz, which they say is a pipeline that “turns the internet into lists of politicians.” It captures snapshots of web pages — plaintext, rendered HTML, and a screenshot kept as evidence — then uses an LLM to extract “human / position pairs” as structured records shaped for ingest into the OpenSanctions data pipeline. The repo ships an AGENTS.md for coding agents.
IRE/NICAR is scraping and archiving police and fire dispatch calls in its hometown of Columbia, Missouri. Cody Winchester’s Playwright scraper runs every 30 minutes via GitHub Actions, appends new incidents to CSV archives in the repo, and sends a Slack alert for each new call. The repo includes a starter Jupyter notebook, which I always like to see in data-journalism projects.
ICIJ released a standalone code formatter that harnesses IntelliJ IDEA’s formatting engine as a command-line tool or embeddable library — no IDE required — with a CI-friendly check mode that fails builds on formatting drift. A few of its recent commits credit Claude.
And the pace of election preparation is increasing: the South Florida Sun Sentinel published the web page code behind its early-voting site maps and candidate guide for Florida’s August 18 primary; Sahan Journal stood up an auto-updating live-results page for the Hennepin County Attorney primary, with a clever bit of code from Cynthia Tu that refreshes a Datawrapper chart using a scheduled notebook run — no server required. And in this update’s most hotly contested race, Charlottesville Tomorrow published a results visualization for the Albemarle County Fair’s “Mayor of the Fair” contest.
By the Numbers
Beyond new repos, 87 news organizations made a combined 4,498 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 | 1,451 |
| Freedom of the Press Foundation | 593 |
| ICIJ | 282 |
| Reuters Graphics | 118 |
| Bloomberg | 117 |
| OpenSanctions | 106 |
| OpenElections | 94 |
| NRK | 83 |
| The Pudding | 82 |
| OpenAleph | 74 |
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.