What it does

This repo is a structured curriculum for learning agentic AI, maintained in Traditional Chinese, Simplified Chinese, and English. It maps 240+ curated projects across eight stages, from LLM basics through prompt engineering, tool use, frameworks, and multi-agent systems. Two tracks split after Stage 2: Track A teaches you to wield CLI agents like Claude Code and Codex; Track B teaches you to build agents from scratch with ReAct, LangGraph, and MCP servers.

The interesting bit

The author treats “awesome list” as a pedagogical format, not a dumping ground. Each stage includes 1-5 hands-on exercises (70-150 lines, dual-path Ollama/Anthropic SDK), estimated time commitments (Track A: 8-10 weeks; Track B: 5-7 months part-time), and role-based branches for researchers, teachers, knowledge workers, and everyday non-coders. The glossary explicitly pairs Chinese understanding names with English technical terms so readers can navigate upstream documentation without getting lost in translation.

Key highlights

• 240+ curated projects with star ratings, target audience, and run instructions (including local LLM options: Ollama, llama.cpp, LocalAI, MLX)
• 23 exercise folders with mock-based tests and dual-path SDK comparisons
• 65 MCP/Skill catalog entries covering the Chinese AI ecosystem (DeepSeek, Zhipu, Kimi)
• Two shared “hubs” — Stage 5 (Claude Code ecosystem) and Stage 8 (Computer Use / Browser Use / Sandbox) — taught from different angles for each track
• A 7-step walkthrough building the same Paper Summary Bot from Stage 1 through Stage 7 (~350 lines)

Caveats

• The repo is curation and documentation, not a framework; you clone it to read the markdown stages, not to pip install anything
• Estimated timelines assume 5-8 hours per week; the author notes the field moves fast enough that 2026 still brings monthly new models and frameworks

Verdict

Worth bookmarking if you’re a Mandarin- or English-speaking developer who wants structured progression rather than drowning in AI Twitter hype cycles. Skip it if you already ship multi-agent systems and just need API docs — this is a syllabus, not a shortcut.