Prompt Engineering

This module is most useful when treated as systems design for model interaction, not as a bag of prompt hacks. The repo currently focuses on the practical subset you can actually apply in production workflows.

Actual Module Contents

1. 01_START_HERE.ipynb
2. 02_basic_prompting.ipynb
3. 03_chain_of_thought.ipynb
4. 04_react_prompting.ipynb
5. 05_structured_outputs_dspy.ipynb
6. 06_long_context_strategies.ipynb
7. 08_assignment.md

What To Learn Here

• How prompt quality changes task reliability
• When few-shot examples are worth the extra tokens
• How ReAct changes model behavior when tools are involved
• Why structured outputs are better than post-hoc parsing
• How to handle long contexts without blindly stuffing documents into prompts
• How reasoning models change prompting strategy and output expectations
• When prompt caching, context compression, and schema-first design matter more than clever wording

Recent 2026 Prompting Topics

• Reasoning-model prompting for deliberate models that think longer before answering
• Context engineering: prompt layering, retrieval packing, and instruction hierarchy
• Prompt caching and prefix reuse to reduce cost and latency
• Structured generation with JSON schema constraints and tool-first interfaces
• Eval-driven prompt iteration instead of subjective prompt tweaking

Recommended Study Order

• Start with 02_basic_prompting.ipynb
• Move to 03_chain_of_thought.ipynb for reasoning patterns
• Study 04_react_prompting.ipynb before Phase 15 agents
• Use 05_structured_outputs_dspy.ipynb to connect prompting with reliability
• Finish with 06_long_context_strategies.ipynb before deeper RAG and agent work

Learning Advice

• Prompting is not a substitute for retrieval, fine-tuning, or evaluation.
• If a task fails, ask whether the real issue is missing context, poor tools, or missing schema constraints.
• Keep examples and output formats explicit; ambiguity is expensive.

Good Follow-On Projects

• A structured extraction pipeline with validation
• A multi-step research assistant using ReAct
• A long-context summarizer that compares chunk-then-summarize vs direct prompting

What Comes Next

• Continue to ../15-ai-agents/README.md if you want tool-using systems.
• Revisit ../08-rag/README.md if you want to improve retrieval prompts and answer synthesis.
• Continue to ../16-model-evaluation/README.md if you want to measure prompt quality instead of iterating by intuition.