MIT researchers propose a new model for legible, modular software (news.mit.edu)

MIT CSAIL researchers led by Daniel Jackson and Eagon Meng introduced a structural pattern for “legible” software that breaks systems into explicit, high-level pieces called concepts and synchronizations. Concepts encapsulate a single coherent feature (state + actions) — e.g., sharing, liking, commenting — while synchronizations are a small domain-specific language (DSL) used to declare how concepts interact. Presented at SPLASH, their paper shows this approach reduces “feature fragmentation” where one feature’s logic is scattered across services and buried in low-level code. Technically, synchronizations are declarative contracts that can be analyzed, verified, and reliably generated by LLMs, enabling clearer semantics, centralized concerns (error handling, storage, formatting), and safer automated code generation. A case study re-factored social features into single concepts with synchronizations describing interactions, making behavior easier to locate, test, and reason about. The pattern also scales: synchronizations can coordinate distributed replicas, support eventual-consistency semantics, and be cataloged as reusable domain libraries. By making inter-component connections explicit, the model promises a practical way for AI-assisted development to propose changes with fewer hidden side effects, and to move software design toward higher-level, human-centered abstractions.