Open-source book 'Understanding AI Agents' hits 3.0 with rewritten chapters and new case studies
yangyi · x · 2026-11-23
Bojie Li's open-source book Understanding AI Agents: Design Principles and Engineering Practice (born from a bootcamp course taught at Turing) has released version 3.0. Built around the core formula Agent = LLM + Context + Tools across ten chapters covering context engineering, memory/RAG, tool design and MCP, coding agents, and agent evaluation.
Key 3.0 updates:
- Most chapters rewritten, shifting from survey-style to teaching-style prose
- New sections on Jev-style decision models for event filtering, tool safety judgment, eval rubrics, RAG filtering, and multimodal information sparsification
- New case studies: personal agents like Meta Muse, Pine, and dsh superagent; office agents like Multica and LobeHub; research agents like Anthropic's Riemann proof and OpenAI's Millennium Problem work
- Revised voice interaction chapter, many redrawn illustrations
- Merged 30+ community PRs, resolved 20+ issues, added a Mongolian translation
The PDF is free on GitHub; feedback via issues is welcome.
More from coding & agent
- Sentient OS (YC F26) launches an open-source personal AI agent that runs entirely on your Mac, no cloud bill — Scobleizer · 2026-11-23
- asc api schema checks catch duplicate flags before requests, exiting 2 — rudrank · 2026-11-23
- If AI proves Riemann, formal verification becomes the infrastructure of all software — inductionheads · 2026-11-23
- LangChain CEO amplifies bold take: LangSmith as the 2027 agent platform endgame — hwchase17 · 2026-11-23
- Codex Computer Use refuses to touch login pages and says the restriction can't be bypassed — Myfirstreddit124 · 2026-11-23
- Developer wires a receipt printer to a Grok bot that prints whenever he's summoned on X — mattyp · 2026-11-23