Applying the scientific method to how we learn, build, and trust research software, now in the age of AI.
Research software runs modern science, and more and more of it is written with AI. Yet most of what we know about doing it well lives in anecdotes, habits and individual projects. Collegeville treats software practice as something to study: we form hypotheses, test them, and share what holds up.
For more than 25 years, senior capstone students at Saint John's University and the College of Saint Benedict have worked as research teams, learning computer science by doing research. That work now explores how AI changes the way students learn, build and judge software.
Scientific software comes out of small teams, shared libraries and large ecosystems. We study which practices deliver real value, how trust is earned through testing and reproducibility, and how AI tools are reshaping development, review and maintenance.
An archive of human skills, AI-assisted workflows and agent skills, grown out of the capstone course.
Visit CorpusCollegeville grew out of decades of teaching and of leading scientific software efforts, including the Collegeville Workshops on Scientific Software and the US DOE Exascale Computing Project. Browse the archived Collegeville site