Stanford CS329Z releases all materials: homework 1 builds a Claude Code from scratch
jyangballin · x · 2026-11-11
Stanford's new course CS329Z 'Engineering AI Agents' (Diyi Yang, Michael Ryan, John Yang) has opened all materials: lectures go live daily on the Stanford Online YouTube channel, and both homework repos are MIT-licensed on GitHub.
Homework 1: build an agent harness from scratch
- No agent frameworks allowed — only a raw interface: messages in, text out
- First half: email prioritization, daily digests, hand-written BM25 retrieval, multi-hop QA over a real corporate email archive
- Second half: an internal assistant for a fictional energy company — design your own tool-call format, doc retrieval, sandboxed terminal, context compaction, cross-session memory, user approval gates
- Grading uses an LLM-simulated user plus an LLM judge; 48 of 88 points come from model-free tests, so self-learners can iterate for free
The course philosophy: build RAG, tool use, and agent loops from scratch, then see how frameworks like DSPy abstract them. Lecture 1 already has 90k+ views.
More from coding & agent
- Engineer hid a pirate-talk prompt injection in his LinkedIn — latest SOTA models now spot the trap — smlpth · 2026-11-11
- Fable 5.2 missing from Grok Bot: Squad founder warns against single-provider AI stacks — tibo_maker · 2026-11-11
- Eric Schmidt: found an agentic AI company now—the window won't stay open — NandoDF · 2026-11-11
- A full org chart of AI skills turns Claude into a one-person company — Med1_Ai · 2026-11-11
- AI makes swapping vendors trivial: your MCP server is one prompt away from replacement — FinanceYF5 · 2026-11-11
- NVIDIA Research Korea hiring agentic AI research scientists, partnering with KAIST — hyunw_kim · 2026-11-11