A 12-week course to help CS graduates systematically understand the LLM field.
60 articles covering everything from neural networks to Transformers, RLHF, AI Agents, RAG, and more.
- CS graduates who've been out of school for a few years
- Have programming skills, understand basic algorithms
- May have learned classical ML (linear regression, SVM era)
- Want to understand "why ChatGPT works", not just use it
| Phase | Weeks | Topics |
|---|---|---|
| Fundamentals | 1-3 | Neural Networks → Transformer → Training (RLHF, DPO) |
| Deep Dive | 4-6 | Inference → Capabilities & Limits → Advanced Architectures (MoE, Mamba) |
| Agentic AI | 7-9 | Agent Basics → RAG → MCP → Reliability & Safety |
| Practice | 10-12 | APIs → Deployment → Applications → OpenClaw Implementation |
See full curriculum: English | 中文
| Day | Topic | EN | 中文 |
|---|---|---|---|
| 1 | Neural Network Overview | Read | 阅读 |
| 2 | Rise and Fall of RNN | Coming soon | 即将发布 |
| 3 | Birth of Attention | Coming soon | 即将发布 |
| 4 | Transformer Architecture | Coming soon | 即将发布 |
| 5 | Encoder-Decoder to Decoder-only | Coming soon | 即将发布 |
| Day | Topic | EN | 中文 |
|---|---|---|---|
| 6 | Pretraining vs Fine-tuning | Coming soon | 即将发布 |
| 7 | Tokenization | Read | 阅读 |
| 8 | RLHF & DPO | Coming soon | 即将发布 |
| 9 | Scaling Laws | Coming soon | 即将发布 |
| 10 | Training Infrastructure | Coming soon | 即将发布 |
More articles added daily (weekdays).
- 15-20 minutes per article
- ~1.5 hours per week
- 12 weeks total
Each article includes:
- Core concepts with intuitive explanations
- Code examples (PyTorch)
- Visualizations and diagrams
- Math derivations (optional, for those who want deeper understanding)
- Further reading and reflection questions
CN-EN terminology reference for key terms used throughout the course.
Topics are prioritized based on actual search trends (March 2026):
| Topic | Trend |
|---|---|
| 🔥🔥🔥 AI Agent | Explosive growth |
| 🔥🔥🔥 What is LLM | Sustained high |
| 🔥🔥 Agentic RAG | +300% |
| 🔥🔥 MCP | Rising |
| 🔥🔥 Transformer | Steady growth |
Found an error? Have a suggestion? Feel free to open an issue or PR.
MIT License - Feel free to use for learning and teaching.
Generated with ❤️ by OpenClaw