The Signal-to-Prototype Loop
A repeatable methodology for turning domain expertise into working AI prototypes — developed over 10 years, stress-tested with 200+ students, documented with evidence.
The Three Tools
Every participant already has three tools. The methodology teaches how to combine them.
Your Brain
Domain expertise from years of training and practice. You understand the problems, the workflows, the users. This is the scarce resource — AI can’t replicate it.
Design Thinking
User-centered methodology for structuring insight into action. Journey mapping, jobs to be done, empathy — the bridge between “this is frustrating” and “here’s what to build.”
Generative AI
Large language models as build partners for rapid prototyping. Not replacing expertise — amplifying it. Claude for thinking and synthesis, Lovable and Claude Code for building.
The Loop
Five steps, repeated daily. Each cycle adds a new design lens. Richer planning produces richer builds.
Layered PRD Builds
Students generate a fresh build every day from the enriched PRD — not iterating on yesterday’s build. By Day 5, five versions of the same app reveal how planning depth compounds.
Why rebuild from scratch instead of iterating? Because comparing daily builds makes the planning-to-building relationship concrete. Students see directly: a richer PRD produces a richer app. The build is disposable. The thinking is cumulative.
Evidence
This methodology was developed over 10 years across 9 cohorts and 200+ students at the UVA Medical Design Program. The February 2026 Health Design Sprint stress-tested it in a two-week intensive format, producing 5 production-quality healthcare applications. The entire process is documented: 519 indexed transcript segments, video recordings, artifact collections, and AI-generated course narratives.