A Dark Information Theory of Certified Discovery

Certified discovery is the practice of deciding absence: a claim stands because a calibrated battery of tests examined it and found nothing. Every such battery is an architecture, and every architecture has a blind region whose size is exact: the largest signal mass a deterministic planted structure can carry while every test passes at its stated level. This paper calls that quantity the dark mass of the architecture and founds its quantitative theory. The founding law is a strong duality (Theorem GD-ES): over block-anchored pools under local coupling, with selection read from the observed spectrum and a member-matched battery, dark mass equals, to first order, the value of a deterministic cap program read off the architecture alone, and no identifiability is assumed. The law has two elementary halves: undetection forces every cap to its bar (Lemma CI); margins on the caps force passage (Lemma SM, Theorem MS). At the closed idealized architecture the value per loaded atom takes a closed envelope form in the amplitude budget and the scan calibration alone (Theorem F'); under the identifiability hypothesis I2' it is min(B, h)^2 exactly at first order (Theorem F). One identification then organizes the theory: concealment from a defender class is keyed covert communication with that class as warden (Theorem T0), and calibration at multiplicity makes the concealable mass linear in the support size where the square-root law of covert communication stood. The exact boundary of the developed theory, and the programs open beyond it, are stated as such.