talks

The CS Curriculum in the Age of Coding Agents

Bill Pugh, Professor Emeritus of Computer Science, University of Maryland
pugh@umd.edu

Given in UC San Diego’s Tales of Modern Software Engineering series on October 12, 2026. An earlier version was given to a class at the University of Maryland on September 29, 2026.

The slides, except for the ones about revising the CS curriculum, will be presented in the “Tales of Modern Software Engineering” series. The slides on revising the CS curriculum will be presented and discussed in the UCSD CS Education Research seminar.

Materials

• Slides, with the references for each slide (HTML; use the Play button for a slideshow)
• Slides as a PDF
• References: every source cited on a slide, with slide numbers

Abstract

At the companies furthest along, most source code is now typed by AI, and nobody reads every line. What is left for the humans is deciding what to build, designing systems that survive change, and knowing whether what got built is what was meant. Those used to be skills you picked up in senior courses, internships, or your first year on the job. They are becoming what it takes to get the first job, and new hires are taking on responsibilities traditionally associated with more senior positions, such as senior engineers or program managers.

I’ll discuss several topics, with many unanswered questions.

• How is software development changing, and which changes will last? Adoption is not uniform, and few people will hazard a guess at where the leading edge will be in two years. How do we pick a target to prepare for?
• What skills should CS students have when they graduate? There is evidence that some of these skills, such as communication and working with people who are not software developers, have not been well covered by CS curricula.
• What do we know about how AI affects learning? There is ample evidence that mental challenge and struggle are required to acquire skills in many domains. As we adopt AI within the CS curriculum, how do we make sure students still encounter the struggle needed to build the skills they will need?
• How should the CS curriculum change, from the first programming course to the capstone? To make room for new material, we will have to either drop some existing topics or teach them much more efficiently.

And the two most urgent questions:

• What advice can we give current CS students?
• How should we change our intro CS courses? Even if we act urgently, a redesigned intro course will not reach students who graduate before 2032.

About the speaker

Bill Pugh is Professor Emeritus of Computer Science at the University of Maryland. He worked on Skip Lists, the Omega Project, the Java Memory Model, FindBugs, and Marmoset, a system for managing student programming project submission, testing, and code review. He has worked at both Google and Apple. He officially retired from UMD in 2012, but re-engaged with the department in 2025 to help figure out how CS curricula should be updated to reflect the impact of generative AI on software development.

Comments and corrections are welcome at pugh@umd.edu; please refer to slides by number.