Answers are grounded in this paper and cite their sources.
Any text that cites figures from code, data or experiments can drift from its source, whether it is written by a person or an AI. This report describes a method for keeping every figure tied to its source: a registry with one row per figure, and a verifier in CI that checks that every figure registered as checked matches its artifact. The case is my Polar Projector paper: the verifier checks the registered figures, but not whether the code does what the text describes, as happened with figures that passed it and were later retracted. As a pilot, the report was written with Actuarii, my governance kit for coding agents, and it counts which writes went through its channel. The method gives traceability, not reproducibility.
No takes yet. Share an insight, caveat, or question.
Alexis Zapico Fernández (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: