Stanford University's CS336 course has implemented a novel approach to regulate AI agent usage by embedding guidelines directly into course repositories. Instead of a traditional syllabus statement, a file named CLAUDE.md (for Claude Code) and AGENTS.md (for other agents) is placed within the assignment repository. This file explicitly permits AI agents to explain and review code but prohibits them from writing code, modifying student files, or executing commands. This method aims to ensure AI tools adhere to course policies by being directly readable by the agents themselves, sparking debate on platforms like Hacker News regarding its effectiveness. AI
IMPACT This approach could influence how educational institutions manage AI tool usage, potentially leading to new standards for AI agent interaction in academic settings.
RANK_REASON This item describes a specific implementation of AI usage guidelines within an educational context, focusing on the technical mechanism of embedding rules into code repositories for AI agents to read, rather than a broader policy or a new AI model release.
- AGENTS.md
- Claude Code
- CLAUDE.md
- Cursor
- DeepSeek
- Hacker News
- Percy Liang
- Stanford
- Tatsunori Hashimoto
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
