Stanford CS336 Guide for AI Coding Assistants
The Stanford Computer Science department has released guidelines for artificial intelligence (AI) coding assistants used in their CS336 course. The goal of these agents is to help students learn through explanation and guidance, not by completing assignments directly.
According to the guidelines, AI agents should provide feedback on code written by students, suggesting improvements and edge cases rather than giving direct solutions. Agents are also encouraged to explain error messages from various tools and programming languages, helping students understand approaches and algorithms at a high level. Additionally, AI assistants are advised to engage in active dialog with students, guiding them through the learning process.
To ensure that students learn by doing and not just watching an AI generate solutions, the guidelines outline specific dos and don'ts for AI agents. Agents should provide general feedback, point students to relevant resources, and explain concepts when needed. They should also help debug code by asking guiding questions rather than providing fixes. On the other hand, agents are prohibited from giving direct implementations, completing TODO sections in assignment code, or editing code in student repositories.
By following these guidelines, AI coding assistants can effectively support students in CS336 while maintaining a learning-oriented approach.
