How Stanford Teaches AI-Native Engineering with Coding Agents
Stanford's CS146S course has adapted to the prevalence of coding agents by shifting homework assignments to require student contributions to unfamiliar open-source repositories, as discussed by Mihail Eric. This educational shift addresses the reality that traditional coding homework can now be easily solved by AI in a single prompt, forcing computer science education to focus on higher-level system understanding and collaboration with AI. In its second iteration, the course collaborates directly with open-source maintainers, challenging students to learn a new codebase, propose changes, and navigate real-world software maintenance alongside AI tools.
## BACKGROUND
Coding agents are AI-powered assistants that integrate into development environments to autonomously handle end-to-end tasks like code editing, testing, and debugging. AI-native engineering represents a paradigm shift where AI is deeply integrated into software development workflows and architecture rather than just being used as a superficial add-on.