Phase 8 · 2–2 weeks
Prompt Engineering & Reasoning
From ad-hoc prompting to a real vocabulary — CoT, self-consistency, budget forcing, reasoning models.
What this phase teaches
Zero/few-shot, CoT, Self-Consistency, ToT, Graph of Thoughts, ReAct, Reflexion, Plan-and-Solve, PAL. Structured output (JSON mode, grammars, constrained decoding). Reasoning-model prompting for o1/o3/DeepSeek-R1. Lilian Weng’s “Why We Think”.
Anchor resources
- DeepLearning.AI ChatGPT Prompt Engineering for Developers
- Anthropic’s Prompt Engineering Interactive Tutorial
- HF LLM Course chapter 12
- Papers: 2201.11903 (CoT), 2210.03629 (ReAct), 2501.19393 (s1), 2501.12948 (DeepSeek-R1)