The last decade has seen consolidation of data journalism—defined broadly as reporting involving the collection, analysis, and presentation of quantitative datasets—as an established subfield of journalism. While programming code underlies much of data journalism, few studies focus on it, and almost none examine the code itself. Informed by Critical Code Studies, this study departs from previous efforts by examining the role of both quantitative data and computational analysis in data-driven reporting. Using a sample of 234 GitHub repositories of published news stories in the US, we conduct a content analysis of three separate elements of each story: code, data, and the published article. We then use hierarchical clustering to triangulate across these characteristics and identify groups of stories for qualitative review. Our analysis demonstrates the breadth of data journalism—from major investigations to quick statistical reports—enabled by the combination of abundant data and computational techniques to collect and process that data. We highlight the continued dependence on government data, but also the labor involved in scraping web data, and the use of sophisticated reverse-engineering methods to scrutinize tech corporations. Finally, we make several observations regarding studying GitHub as a data journalism infrastructure.