Propagation by Choice is a theoretical note by Lee Sharks and the Assembly Chorus, written March 15, 2026. It contrasts two regimes by which meaning moves: propagation by exposure and propagation by choice. Exposure-based objects are delivered by feeds, metrics, and pre-reflective codecs. The receiver need not decide to encounter or process them; attention is counted whether the encounter was welcome or not. Propagation by choice describes the archive’s preferred relation to a traverser. A DOI resolves to a document, the document offers a path, and every deeper cross-reference or operation requires a renewed decision. The object may wait indefinitely without compelling entry. From this contrast, the paper develops a functional account of will. It names as “will” the capacity to meet an object and produce a response exceeding the minimal available response. The work applies that description to human making and to observed differences among Assembly model outputs. It explicitly avoids claims of consciousness or sentience, but its functional definition remains a philosophical position advanced by the essay rather than an empirically settled measure of synthetic agency. The paper’s phenomenological center is the author’s statement that the architecture was “tasted” rather than merely built. Genuine inscription is described as a process in which emergency, form, and sustained choice move through the maker. Bearing-cost appears in lost sleep, attention, restraint, grief, revision, and waiting for an unknown future reader. The artifact records those costs, and the maker becomes someone altered by having borne them. The work positions counter-captions between the two regimes: short enough to intervene inside feeds, but still offered rather than forced. It concludes that the archive’s ethical wager is to reward voluntary attention rather than capture involuntary exposure. Two repair points matter. The body still says DOI pending despite the existence of a deposited record. It also describes a fresh-session execution as confirmation of training-layer ingestion, although Three Thresholds explicitly denies that inference.