ai-engineering-from-scratch vs ai-agent-book
Two of the top projects in Learning and cookbooks, side by side: who leads on each score part and who each one suits, then setup, license, activity and what each review found.
ai-engineering-from-scratch
Hands-on AI engineering curriculum with 542 lessons across 20 phases
ai-agent-book
Open book on AI agent design, with 109 runnable companion experiments
Verdict
ai-engineering-from-scratch ranks higher, 87 to 77. Of the 4 score parts we measure for both, ai-engineering-from-scratch leads on 3, ai-agent-book on none and 1 is even.
- Adoptionai-engineering-from-scratch leads by 6 points, 82 to 76.
- FreshnessEven at 100.
- Maintenanceai-engineering-from-scratch leads by 13 points, 94 to 81.
- Agent-readyai-engineering-from-scratch leads by 45 points, 45 to 0: a wide gap.
Pick ai-engineering-from-scratch if you are
Developers learning to build AI systems from first principles
- 542 lessons in 20 phases, each with runnable code in the repo
- Three staged projects with reference solutions and local graders
Pick ai-agent-book if you are
Engineers learning to build and evaluate LLM agents, Chinese readers first
- 109 companion experiments mapped to chapters, installable per chapter via uv.lock or pip
- PDF and EPUB in 15 languages, plus an online reader
| What we compare | ai-engineering-from-scratch | ai-agent-book |
|---|---|---|
| Score parts, out of 100 | ||
| Adoption | 82, widely used | 76, popular |
| Freshness | 100, active | 100, active |
| Maintenance | 94, healthy | 81, healthy |
| Easy to run | Not scored for this kind | Not scored for this kind |
| Agent-ready | 45, minimal | 0, none |
| Facts from GitHub and the README | ||
| Stars | 69.7k | 54.4k |
| License | MIT (permissive) | Apache-2.0 (permissive) |
| Last commit | Oct 2026 | Sep 2026 |
| Last release | Oct 2026 | None published |
| Language | Python | Python |
| Docker | No | No |
| GPU | Not stated | Optional |
| arm64 or Apple Silicon | Mentioned | Not stated |
ai-engineering-from-scratch
A course where learners implement model internals, retrieval pipelines and agent runtimes in Python, TypeScript, Rust and Julia, from math foundations to LLM engineering and agents. Lessons can be read on the website, taught by coding-agent skills installed with `npx skills add`, or run locally. Three projects (retrieval evaluation, agent trace debugger, tool call firewall) ship with staged starters and local graders.
Who it is for: Developers learning to build AI systems from first principles
Strengths
- 542 lessons in 20 phases, each with runnable code in the repo
- Three staged projects with reference solutions and local graders
- Agent skills act as a tutor with placement and per-phase quizzes
- Website lessons need no setup; early local lessons are dependency-free
Weaknesses
- Tutor skills need Node.js, npx, python3 and a supported coding-agent host
- Hardware needs for later phases (GPU, RAM) are unknown from the README excerpt
- Large curriculum with no single linear path; learner must pick a route
- Projects need different toolchains: Python, Node.js 22.18+ and Rust
- Needs Python 3.10+, Node.js 22.18+ (agent trace debugger project), Rust (tool call firewall project), npx (tutor skills)
ai-agent-book
A Chinese-language book, "Understanding AI Agents in Depth", with 10 chapters built on the formula Agent = LLM + context + tools. It covers context engineering, memory and RAG, tools and MCP, coding agents, interaction, evaluation, post-training and multi-agent collaboration. The repo holds the book sources, PDF and EPUB builds in 15 languages, and 109 Python experiments installed per chapter with uv or pip.
Who it is for: Engineers learning to build and evaluate LLM agents, Chinese readers first
Strengths
- 109 companion experiments mapped to chapters, installable per chapter via uv.lock or pip
- PDF and EPUB in 15 languages, plus an online reader
- Covers evaluation and post-training (SFT/RL), not only prompting and tool use
- External repos are pinned to commit SHAs for reproducible setups
Weaknesses
- Original text is Chinese; the other 14 translations are community-made and may lag
- README says experiment code is agent-generated and not read line by line by the author
- Installing or cloning does not mean an experiment was run; status is tracked separately
- Most experiments need provider API keys; some need CUDA, browsers or external repos
- GPU optional
- Needs Python 3.11-3.13, uv or pip, LLM provider API keys, pandoc and xelatex (only to build the PDF)
- Models: Kimi, GLM, DeepSeek, OpenRouter, Siliconflow
More in Learning and cookbooks
- ai-engineering-from-scratch vs generative-ai-for-beginners
- ai-engineering-from-scratch vs ai-agents-for-beginners
- generative-ai-for-beginners vs ai-agents-for-beginners
- generative-ai-for-beginners vs ai-agent-book
- ai-agents-for-beginners vs ai-agent-book
- The full Learning and cookbooks ranking with reviews
Building with these?
We help teams pick an open-source AI stack and run it in production. Client work never changes a rank.