[One Million AI Engineer Special] How Stanford Teaches AI-Native Engineering — Mihail Eric, Monaco

Mihail Eric, a Stanford instructor and head of AI at Monaco, teaches CS 146S, "The Modern Software Developer," a course that has been refreshed for a second iteration. The genesis of the class came in late 2024, when Eric observed that coding agents were fundamentally changing how developers worked, but the emerging best practices existed only in scattered blog posts. He spent nine months getting the course approved, taught it for the first time in fall 2025, and is now updating it for a world where autonomous coding agents have become dramatically more capable. The syllabus covers everything from building a coding agent in 200 lines of code to the architecture of a full software factory. The through-line is that a well-trained AI-native engineer can be among the most valuable assets at any company, but the floor for entry-level talent has risen sharply.

The curriculum as a practitioner's roadmap

Eric's course is not for beginners; it assumes a solid computer science foundation. The core shift from the first iteration to the second reflects a maturing field. Last year, the class was a survey of every place AI could be applied in the software development lifecycle. This year, principles are solidifying around careful context engineering, design and architecture as a way to handle complexity, and verifiability at scale.