This conceptual preprint introduces Cognitive Aim as the target-layer selection component of intelligent action. It proposes the P/Q/A distinction: Cognitive Power concerns what cognition can process, Cognitive Quality concerns how well cognition is disciplined, and Cognitive Aim concerns what cognition is directed toward. The paper argues that many intelligence evaluations validly measure performance inside supplied frames, but do not by themselves test whether a person, institution, or system can select the relevant target, frame, problem, measure, or layer when that frame is not already supplied. The paper develops the supplied-frame argument, defines the P/Q/A architecture, compares illustrative diagnostic cases, situates the claim in adjacent literatures including rationality research, problem formulation, Type III error, systems thinking, Goodhart-style proxy failure, and AI specification gaming, and states boundaries against psychometric, hidden-cause, and always-upstream overclaims. This paper is one of two linked conceptual preprints. The companion paper, Power, Quality, Aim: The Operational Core of a Layered Atlas of Intelligence, places the P/Q/A operational core within a broader layered architecture of intelligence constructs.