The curriculum

Learn prompt engineering — as a discipline, not a bag of tricks.

Most prompt advice is folklore. This curriculum is built on the empirical work of Sander Schulhoff's Prompt Report and the Learn Prompting framework: the techniques with real evidence behind them, taught with concrete before-and-after examples drawn from the prompts in this library.

19

Lessons

15

Core techniques

~134

Minutes total

Start the first lesson →

PART 1

Foundations

What prompting is, and how a strong prompt is built.

01 What Prompt Engineering Really IsPrompt engineering is the empirical craft of communicating intent to a model — and it is more critical, not le… 7 min 02 The Anatomy of a Strong PromptEvery effective prompt is assembled from the same handful of components — learn to name them, and you can buil… 8 min

PART 2

The Core Techniques

The fifteen techniques that do the heavy lifting — in priority order.

#1 Few-Shot PromptingShow the model two to five concrete examples of the input-output pattern you want instead of describing it in … 8 min #2 DecompositionBreak a hard problem into solvable sub-problems before asking the model to solve any of them. 7 min #3 Self-Criticism & Self-RefinementMake the model critique and revise its own output — but cap the loop at one to three passes before quality ero… 7 min #4 Context & Additional InformationGive the model the background it needs up front — relevant, ordered, and free of noise — so it reasons over fa… 6 min #5 EnsemblingRun a prompt several ways, then let the answers vote — trading compute for reliability on problems where a sin… 7 min #6 Structured Output FormattingMake the model emit output your code can parse — by constraining the format with schemas, tags, and delimiters… 7 min #7 Constraint SettingTell the model what it may not do, how far it may go, and when to stop — constraints are how you make outputs … 6 min #8 Voice & Tone ConsistencyPin down persona, register, and style explicitly — and re-anchor them, because tone drifts over long sessions. 7 min #9 Error Handling & Edge CasesDefine what the model should do when an input is missing, ambiguous, or outside the job — before it invents an… 7 min

PART 3

Advanced Patterns

Reasoning, chaining, and steering models with principles.

#10 Multi-Step Reasoning ChainsChain-of-thought makes non-reasoning models think out loud — but forcing it on reasoning models is often a mis… 7 min #11 Prompt ChainingDecompose a workflow into a pipeline of focused prompts so each stage is reliable, inspectable, and independen… 7 min #12 Meta-PromptingUse the model to write, critique, and refine its own prompts — and treat its output as a draft, not an oracle. 7 min #13 Negative PromptingTelling the model what to avoid is a real tool, but a sharp one — it often works better when you flip the proh… 6 min #14 Constitutional / Principle-Based PromptingGive the model a short set of governing principles and an explicit step to check its own output against them b… 7 min #15 Retrieval-Augmented Prompting (RAG)Ground the model in retrieved evidence so it answers from your data instead of its training-time memory. 8 min

PART 4

From Chat to Production

Separating folklore from findings, and shipping prompts you can trust.

18 What Doesn't Work (and Myths)Separating folklore from findings: the popular prompting tricks that the evidence does not actually support. 8 min 19 From Chat to ProductionA prompt you tweak by hand in a chat window and a prompt that runs ten thousand times a day are two different … 7 min