XNews

Stanford AI Agent Courses for Fall 2026

Stanford AI Agent Courses for Fall 2026

Stanford University announced two Fall 2026 courses focused on AI agents: CS 329Z (Engineering AI Agents) and CS146S (The Modern Software Developer). - CS 329Z covers the full stack from simple LLM pipelines to autonomous agents, structured around decomposition, data, and evaluation. Modules include LLM APIs, RAG, tool calling, frameworks (DSPy, LangChain), design patterns, memory, multi-agent systems, optimization, data flywheels, evaluation infrastructure, safety, coding agents, and proactive agents. Assignments involve building an agent from scratch and designing an evaluation suite. - CS146S, taught by Mihail Eric, focuses on the shift from human-written code to intent-defined, agent-executed development. The 10-week curriculum progresses from understanding agent internals to arming single agents, adapting codebases for agents, and scaling to AI-native teams and software factories. A unique requirement: students must submit real pull requests to 15 partner open-source projects (30% of grade), with mentorship from companies like Browserbase, HeyGen, Semgrep, and Vercel. Guest speakers include creators of Claude Code, Devin, and Replit.

Key insights

CS 329Z Treats Agent Engineering as a Systems Discipline, Not Just Model Prompting

The course frames AI agents as compound systems requiring careful design of decomposition, data pipelines, and evaluation harnesses. By covering frameworks, memory architectures, and safety, it positions agent development as an engineering practice with trade-offs in cost, latency, and reliability, moving beyond single-model interactions.

CS146S Redefines Software Development Around Human-Agent Collaboration

The course's premise is that developers will increasingly define intent and let agents execute, requiring skills in context engineering, agent-ready codebase design, and agentic code review. The mandatory open-source contribution component forces students to apply these skills in real, maintained codebases, developing 'software taste' that cannot be learned in sandboxes.

Industry Partnerships and Guest Speakers Ground the Curriculum in Current Practice

Both courses feature direct involvement from leading AI tooling companies and creators. CS146S's 15 OSS partners and guest list (Claude Code, Devin, Replit, Vercel) ensure students engage with the tools and workflows shaping the 2026 AI development landscape, making the courses a direct pipeline to industry needs.

Source posts