Stanford Teaches AI Engineering Through Open-Source Contributions

Summary

Stanford adjunct professor and Monaco AI lead Mikhail Eric discusses the second iteration of the CS146S course in an AI Engineer podcast episode exclusive to 1 million subscribers. In this ten-week course, students are required to learn unfamiliar open-source repositories, propose patches, and defend these changes to actual maintainers. Eric also covers Monaco's internal agent workspace applications.

Why it matters

This is one of the first initiatives to formalize AI-native development pedagogy at a top-tier university. It moves beyond blog-post-level knowledge by focusing on a structured curriculum that carries real-world responsibility, with open-source maintainers acting as evaluators. Eric’s role as a practitioner at Monaco creates a rare bridge between academic curricula and production agent infrastructure. This offers a useful reference point for how institutions can adapt their computer science curricula as coding agents mature.

Highlights

• Students must learn an unfamiliar repository, propose a change, and explain the patch to the maintainer who can merge it.
• Monaco’s tooling philosophy: 'Buy to learn, build to protect' — this approach is applied to agent workspaces, code review tools, and on-call agents.
• It is emphasized that code review becomes the emerging bottleneck once code generation surpasses human review capacity.

Critical take

Although framed as a deep technical review, the episode is largely a curriculum tour presented within the context of subscriber milestone celebrations and Monaco brand promotion. The text cuts off midway, and the course has only been offered once before. Consequently, claims about 'bypassing contact with their own engineering teams' rely on feedback from a single cohort rather than validated pedagogical outcomes.