The paper turns closed Herculaneum papyri into inspectable surfaces with phase contrast microtomography, explicit meshes, and machine learning used as a visibility aid instead of an OCR oracle. The restraint carries a lot of the argument, since PHerc. 1667 gets bounded coverage criteria while PHerc. Paris 4 gives a separate volumetric ink check. Click for the rare research writeup where the strange premise comes with failure modes and not just victory laps.