The supplementary material is a computer

A sociologist submits a paper. The journal asks for supplementary material: data, code, a README explaining how to rebuild figure 3. Historically that's a zip that works on the author's machine and nowhere else.

This time the footnote is a workspace URL. One panel is the cleaned dataset in SQLite, queryable. One panel is the analysis code in an editor. One panel is a terminal where make figure3 runs. The container was snapshotted the day of submission, so what reviewers open is exactly what produced the results, down to the package versions.

Reviewer two changes a parameter and reruns the regression. Forty seconds. They didn't install R, didn't email the author, didn't guess at a dependency.

Reproducibility stops being a promise in the methods section and becomes a URL in the footnote.