The global push to expand K-12 computing education raises questions about how to integrate computing meaningfully into disciplinary curricula. This doctoral research investigates how programming computational models can support students’ engagement with computing concepts and practices while supporting learning in science and engineering. The work is grounded in computational literacies, which frame computing as a form of reading, writing, critique, and expression, and leverages agent-based modeling (ABM) to help learners reason about complex systems. A central design challenge for K-12 modeling tools is balancing accessibility for novices with opportunities for deeper computational engagement. Domain-specific, block-based ABM environments lower entry barriers but often hide underlying computational structures, limiting learners’ agency. This research explores block unpacking, a feature that allows students to progressively open and adapt the internal logic of pre-built domain-specific primitives, as a way to address this tension. Two studies structure the dissertation. The first analyzes sixth-grade classrooms in California where students programmed diffusion models using MoDa, a domain-specific, block- and agent-based environment. Through learning analytics and qualitative interaction analysis, I characterize students’ design moves and problem-solving strategies during model construction. The second study uses a design-based research approach with Brazilian high school students, co-designing a four-session curricular unit on eutrophication. This unit leverages block unpacking to invite learners to modify computational structures while reasoning about interdisciplinary scientific phenomena. Findings aim to produce empirically grounded design principles for modeling environments and curricular materials that balance usability and depth, supporting computational literacies and the meaningful integration of computing into STEM education.