Stanford launches CS329Z, a new fall course on engineering AI agents
jyangballin · x · 2026-09-13
Stanford is offering a new fall course, CS329Z: Engineering AI Agents, taught by Diyi Yang, Michael Ryan, and John Yang.
The course is premised on the shift from monolithic LLMs to compound AI systems — pipelines combining models, retrievers, tools, and optimizers. Students will cover:
- The full spectrum from simple LLM pipelines to compound systems to autonomous agents
- Three core engineering challenges: problem decomposition, data collection/curation, and building evaluations
- Implementing core components (RAG, tool use, agent loops) from scratch, then seeing how frameworks like DSPy abstract these patterns
- Hands-on practice via two applied homework assignments and a quarter-long project building, optimizing, and evaluating agentic systems
Classes meet twice weekly; lecture materials will be published online. Week 1 readings include the BAIR blog post "The Shift from Models to Compound AI Systems" and Andrew Ng's agentic design patterns series.
Related event: Stanford Launches New AI Courses for Fall, Including AI Agent Engineering(2 posts)→
More from coding & agent
- After 25 weeks: four graphs, no LLM in promotion path, and 3 ways agent memory broke — Conscious_Detail_128 · 2026-09-13
- 16 Calls in One Hour: A Metacognitive Loop That Knows When to Stop Self-Correcting — Dzikula · 2026-09-13
- PyroDash: a 4B model learns when to call a big LLM, cutting cost to 1/30 — 机器之心 · 2026-09-13
- Draw Things launches Local Code beta: local coding agents on Mac at ~980 tok/s prefill — liuliu · 2026-09-13
- Agents get stronger weekly yet still solve the wrong problem: AI takeover ETA? — BarriosA2I · 2026-09-13
- AI assistants fix the crashing line, not the system state that keeps producing it — ClickOk5811 · 2026-09-13