Why prompts matter
The Skill Stack
People say "prompt engineering" as if it's one thing. It isn't. It's four sub-skills that compound. The reason a senior prompter gets a usable answer on the first try, while a beginner needs five rounds, is not magic — it's that the senior is exercising all four at once.
The four sub-skills
| Skill | What it means in one line | Where you'll learn it |
|---|---|---|
| Specifying | Pinning down who, what, where, length, tone, format. | Modules 1–2 |
| Patterning | Knowing which prompt shape (zero-shot, few-shot, chain-of-thought, persona, decomposition) fits the task. | Module 3 |
| Iterating | Repairing a bad output instead of restarting from zero. | Modules 4–5 |
| Designing systems | Writing reusable system prompts, grounding with sources, picking the right model. | Modules 6–9 |
Each skill is independently learnable, and each one alone gets you better outputs. Stacked, they're the difference between "AI gave me something I had to rewrite" and "AI gave me something I shipped."
The four sub-skills, in the order this course teaches them
Pin down who, what, where, length, tone, format
Match the prompt shape to the task
Repair a bad output, do not restart
Reusable system prompts, grounding, model choice
How they reinforce each other
A few examples of the stack at work:
- A vague prompt + the wrong pattern = generic mush.
- A specific prompt + the right pattern, but no iteration skill = you give up after the first weird output.
- All four together = the model becomes a real teammate for a slice of your work.
You don't need to memorise this. You'll feel it as you go through the course. Notice when a lesson is teaching specifying (Module 2: anatomy of a prompt), when it's teaching patterns (Module 3), when it's teaching iteration (Modules 4–5), and when it's teaching system design (Modules 6–9).
What "good" looks like by the end of Module 5
- Write a prompt that names role, task, context, input, and output spec without thinking about it.
- Pick a pattern (few-shot for classification, persona for tone, decomposition for planning) without having to look it up.
- Read a bad output, name why it's bad in one sentence, and ship a corrective second prompt.
- Stop restarting from a blank message every time something goes wrong.
That's a real skill, and it's a daily-work skill. The API stuff (Modules 6–9) is on top of this — but this stack is the part that changes how you actually use AI tools every day.
Next: a tiny exercise — write your first deliberately specific prompt and feel the difference. :::
Sign in to rate