I’ve been traveling and skipped last week, so this is a three-week roundup. In that time, 14 news organizations created or opened 18 public repositories on GitHub. Highlights below.
Highlights
Diagram Chasing published the code and data behind the delightful project “Mapping India’s Clouds”, a daily interactive map of cloud cover over India. A GitHub Actions scraper collects meteograms from roughly 1,245 India Meteorological Department stations each day and publishes the result as Parquet and CSV under an open license. The same outlet opened Argo, the SvelteKit scaffold it uses to spin up new story microsites, which ships a CLAUDE.md and a .claude/commands directory.
Deutsche Welle published the data, code and methodology behind its analysis of “energy leapfrogging” — the finding that some of the world’s poorest countries are building renewable electricity systems from the start, while the West is taking a “fossil fuel detour.” The analysis backs a DW video called, “The unexpected way India is beating the U.S.” The repository carries the full analysis.Rmd, the raw inputs, and source links for each dataset.
ICIJ released Nuditag, a command-line tool that scans a directory of images and video and tags each file as “safe” or “not safe for work.” It runs locally (on the CPU — no expensive GPU needed) and is built for triaging leaked media collections where sending files to a cloud service is not an option. It processes files in parallel, resumes interrupted runs, and writes results to CSV or JSON.
Buried Signals opened Anomaly, a pipeline that finds irregularities in structured data — repeated payments, shared addresses, statistical outliers — and hands them to a reporter to follow up on. The repo ships 28 skills for pulling public records from SEC EDGAR, OpenFEC, CourtListener, OpenSanctions and EU procurement databases. They also opened Splash, a local-first visual journalism tool that turns reporting material into web interactives, scrollytelling and static graphics, keeping source files and approvals in a directory the journalist controls. It is live at splash.buriedsignals.com.
The Baltimore Banner published its subscriber engagement scoring engine, which assigns each reader a weekly engagement score and a behavioral cluster, separating habitual returners from one-time readers. It runs on DuckDB against local files, on a laptop or in CI, with no data warehouse and no vendor SDK, so another newsroom can run it without adopting the Banner’s stack. It is pre-release, with no real interface yet. Eleven of its last 20 commits credit Claude.
Northwestern University’s Computational Journalism Lab put out two things. The first is a collection of agentic AI investigative workflows submitted to its Agentic AI Investigative Journalism Challenge. Winners were announced last week. The second is a dashboard of the federal AI Use Case Inventory (code), a dataset by the Office of Management and Budget covering 3,611 reported AI applications across 41 federal agencies.
CalMatters opened its California campaign finance downloader, a command-line tool for bulk-pulling contribution and independent expenditure data from the Secretary of State’s PowerSearch database. It filters by candidate, office, committee, election cycle or ballot measure, flags problems in what comes back, and ships a data dictionary. It was built for the newsroom’s voter guide. The README warns “This tool is currently a prototype and may not be further developed.”
ABC News (Australia) opened the code behind its visualization of Elon Musk’s wealth, the scrollytelling piece that ran in June when the SpaceX listing made Musk, temporarily, the first trillionaire. The story sizes the number against the next three richest men combined and against Australian wealth figures. (We held this one last month because we could not confirm the story had run; it had.)
Civio opened Butterbot, a minimal Slack agent harness that connects a language model to a workspace and lets it run shell commands, read and write files, and execute reusable scripts it calls skills. It runs over Socket Mode, so it needs no public URL, and sandboxes execution in Docker.
The South Florida Sun Sentinel opened the map behind its District 20 primary coverage, published the day after Florida’s August 18 congressional primaries.
By the Numbers
Beyond new repos, 103 news organizations made a combined 7,075 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 | 1,855 |
| Buried Signals | 618 |
| ICIJ | 547 |
| Freedom of the Press Foundation | 434 |
| MuckRock | 282 |
| ABC News (US) | 270 |
| OpenSanctions | 269 |
| Bloomberg | 191 |
| OpenAleph | 175 |
| OpenStates | 139 |
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.