Palestra

Level 2: The Prompt Engineer

Move from casual conversation to precise instructional design. Learn advanced prompting techniques that maximize AI output quality.

Welcome to Level 2

In Level 1, you learned that AI is an inference engine producing probabilistic predictions. You used three basic patterns -- summarize, draft, brainstorm -- and saw that specificity improves results.

Level 2 transforms you from a casual user into a prompt engineer: someone who crafts inputs that minimize ambiguity and maximize the accuracy and usefulness of model output. Prompt engineering is not about tricks or hacks. It is about understanding how the model processes your instructions and structuring them accordingly.

What you will learn

ModuleTopicCore Idea
2.1Zero-Shot and Few-Shot PromptingTeaching the model by example
2.2Chain-of-Thought ReasoningForcing step-by-step problem solving
2.3Roles and PersonasShaping output through identity assignment
2.4Parameters and ControlTemperature, top-p, max tokens, delimiters
2.5Recursive Self-ImprovementHaving the model critique and refine its own output
2.6CheckpointTest your prompt engineering skills

The shift in competency focus

At Level 1, the emphasis was on Discernment -- recognizing that AI output is probabilistic. At Level 2, the emphasis shifts to Description: the ability to communicate goals and constraints so precisely that the model has minimal room for misinterpretation.

The difference between a Level 1 and Level 2 user is not that the Level 2 user knows more about AI. It is that the Level 2 user writes prompts that reliably produce the output they need on the first or second attempt, rather than the fifth or sixth.

Key Takeaway

Prompt engineering is instructional design for machines. Every ambiguity in your prompt is a dimension where the model guesses. Your job is to remove as many of those guesses as possible through structure, examples, and explicit constraints.

Connect

Start with Module 2.1 on zero-shot and few-shot prompting. You already know zero-shot from Level 1 -- now you will learn how adding examples fundamentally changes what the model can do.

2.1 Zero-Shot and Few-Shot Prompting