When Machines Know Everything: Redesigning Education Around What AI Cannot Replace — An Open Adaptive Competency Education (OACE) Architecture Integrating Reliance-Aware Assessment into Competency-Graph Progression

The rapid, near-universal availability of generative artificial intelligence has quietly dismantled a load-bearing assumption of modern education: that the ability to produce a correct answer, unaided, is a reliable proxy for competence. When a free chatbot can solve, explain, and format the majority of tasks assigned in a conventional grade-ladder curriculum, the traditional coupling of assessment to information recall and reproduction becomes unstable. This paper argues that the appropriate response is neither to ban AI from education nor to treat it as an unqualified substitute for instruction, but to redesign the underlying architecture of how learners progress and how that progression is verified. We introduce Open Adaptive Competency Education (OACE), a design-science artifact that reorganizes learning around a competency graph rather than a fixed age-grade-subject ladder, and that embeds a reliance-aware progression mechanism: a formal logic by which a learner's documented pattern of tool dependence — measured using constructs adapted from the emerging AI-reliance and cognitive-offloading literature — becomes an explicit input into whether, when, and how a learner advances to the next competency node. This integration is, to our knowledge, novel: existing work independently addresses competency-graph curriculum structures, AI-reliance measurement, and assist-versus-test transfer effects, but no prior architecture connects reliance measurement directly to progression logic within a single implementable system.