Digital Media: From Prototypes to Products and Services

This is a practical software engineering course on building a minimum viable product or service from an interactive prototype for a mobile or web application, using agentic artificial intelligence (AI) development tools to make professional-grade software engineering accessible without requiring a traditional programming background. We begin with a detailed review of students' designs, with particular attention to the underlying design system and component library, translating these into a clear technical specification that an AI coding agent can act on. Rather than writing code from scratch, students work as technical directors—specifying, evaluating, and refining the output of agentic tools like Claude Code that can autonomously scaffold, build, debug, and iterate on a codebase. The first half of the course is dedicated to building a fully functioning demo of your application, where the front-end user experience is powered by your component library and the backend is implemented on Convex—a real-time backend platform that pairs particularly well with agentic development workflows given its code-first approach to data, functions, and sync. AI-powered features are built using Convex's native Agent Component, which manages threads, messages, and multi-step workflows in the same TypeScript environment as the rest of the application, eliminating the need for a separate AI software development kit (SDK). Students develop sufficient technical fluency to direct AI agents effectively, understanding the architecture of a modern web application, reading and evaluating generated code, and knowing when and how to intervene. The second half of the course offers two paths depending on your goals. The first path focuses on enriching your demo with novel features—iterative refinements to interaction and interface design, integration with third-party services, real-time interactivity, and working with cloud storage and live data. The second path focuses on preparing your application for actual launch—covering deployment, security, and scaling considerations for a production-ready product or service. Throughout both halves, AI agents are not just tools for writing code—they are collaborators in problem-solving, debugging, and architectural decision-making. Learning to work effectively with agentic systems is itself a core skill this course develops. The work in this course draws on a mix of the following tools and technologies: Claude Code, Notion, Framer, Next.js, Convex, GitHub, and Visual Studio Code.