Specifications, Tests, and Audits: the Building Blocks of First-Year CS in an Agentic Era

Deep and interesting questions sit underneath the well-worn question of “should we ban or embrace AI in first-year CS”. Early CS curricula have long embraced assumptions that now warrant reconsideration (e.g., that we should emphasize writing code over reading it). This talk highlights several assumptions and makes a concrete proposal: first-year computing courses can and should emphasize core software-engineering practices—such as specification, property-based testing, and code audits. We’ll explore this idea through both research literature and promising results from two experimental courses (one last spring for students with some CS background, and one this semester for novices). Ultimately, we hope to provide a research-based foundation for discussions about the affordances and pitfalls of agents in early computing education.