Where the Boundary Moves: Cognitive Bottleneck Migration, Framework Construction, and Subject Continuity in Artificial Intelligence

Artificial intelligence is progressively acquiring capabilities once treated as defining featuresof human intelligence, including calculation, symbolic reasoning, natural-language generation,scientific question generation, long-horizon information retention, and increasingly complexinteraction with the environment. As these capabilities become engineered and abundant, theapparent boundary between human and artificial cognition repeatedly recedes. This paperargues that the phenomenon should not be understood merely as a sociological tendency tomove the goalposts of “intelligence.” It reflects a more structural change: when a cognitivefunction becomes abundant, inexpensive, and automatable, the principal bottleneck in knowledgeproduction and cognitive agency tends to migrate toward higher-order functions that organize,select, evaluate, or reconstruct that function.The paper develops a unified moving-boundary framework and makes four connected conceptual contributions. First, it proposes a principle of cognitive bottleneck migration: automation atone cognitive layer tends to shift relative scarcity toward complementary meta-level functions.Second, it represents a cognitive framework asF =(O,R,I,E,S)and defines the corresponding question space Q(F), allowing a distinction among withinframework problem solving, framework stress, and genuine framework reconstruction thatalters what can count as a well-formed question. Third, it introduces a conceptual provenancedecomposition,Π=(G,A,I,X,T),distinguishing origination, articulation, inference, extension, and transformation, and derives theprinciple that solvability does not establish conceptual ancestry. Fourth, it develops a multi-layercontinuity framework for artificial cognition, separating memory, informational, relational, andfunctional continuity from ontological continuity and from continuity of a first-person point ofview.The paper further distinguishes lived and inherited provenance of knowledge, and introducesagent-situated retrieval, salience structure, and endogenous cognition as analytic dimensions fortracking the development of increasingly subject-like artificial systems. It does not claim thatcurrent AI systems are conscious, nor that increasing functional sophistication must producesubjective experience. Rather, it explicitly preserves the gaps between subject-like organization,1ontological identity, and first-person experience as distinct questions, the last of which presentlylacks a decisive discriminator.The central claim is therefore dynamic rather than exceptionalist: the theoretically relevanthuman–AI boundary should not be attached to any permanently privileged task. The relevantquestions are where the boundary lies now, why it lies there, and what evidence would justifysaying that it has moved.