Learn Prompting that works
Beginner Free
Prompting that works
A practical, model-agnostic approach to prompt engineering: how to structure a prompt, when examples help, how to get structured output you can parse, how to test a prompt before you ship it, and the cost/latency trade-offs behind your choices.
- 5 lessons
- 31 min total
- Quiz in every lesson
What you'll learn
- A well-structured prompt separates role/instructions, context/data, the actual task, and output format into distinct, labeled sections.
- Few-shot prompting (showing 1-5 example input/output pairs) works especially well for format and tone, where 'show, don't tell' beats a written description.
- Asking a model to 'return JSON' in plain instructions is not reliable on its own - it can add prose before/after the JSON, use the wrong field names, or produce invalid JSON entirely.
- A prompt is code that produces probabilistic output - it needs the same discipline as any other code path that handles real inputs: a test set, not a vibe check.
- You pay for both input and output tokens, and output tokens are typically priced higher - a verbose system prompt on every request adds up fast at scale.
Curriculum
- 1. Structure The four parts of a well-built prompt, and why order and separation matter more than clever phrasing. 6 min (completed)
- 2. Examples and few-shot prompting When showing examples beats describing rules, how many to use, and the trap of examples that are too similar to each other. 6 min (completed)
- 3. Structured output Getting JSON you can actually parse: schema-constrained decoding vs prompting alone, and what to do when a provider doesn't support the former. 7 min (completed)
- 4. Testing prompts Why a prompt that 'looks right' twice isn't tested, temperature 0 doesn't mean deterministic, and how to build a minimal prompt test suite. 6 min (completed)
- 5. Cost and latency Where your token spend actually goes, streaming vs waiting for the full response, and the cheap wins before you reach for a smaller model. 6 min (completed)