COURSE 08L0100% FREE
Verified 2026-08-10

Prompting

A **prompt** is the text (and sometimes images) you give a model so it can continue. Because LMs predict next tokens, your words are the steering wheel—not a magical spell language.

What it is

A prompt is the text (and sometimes images) you give a model so it can continue. Because LMs predict next tokens, your words are the steering wheel—not a magical spell language.

Why it matters

Clear prompts reduce garbage. They do not guarantee truth. Prompting is a skill of specification + verification, not “secret jailbreak art” (see Course 29 literacy).

How it works (plain)

Helpful patterns:

  • State the goal
  • Give context the model cannot see otherwise
  • Specify format (bullets, table, JSON-ish structure)
  • Provide examples when style matters
  • Ask it to check or list uncertainties

Then you verify with sources or tools.

Everyday example

Bad: “Write about marketing.” Better: “Write 5 email subject lines under 50 characters for a lawn-care spring discount in Florida; friendly tone; no fake scarcity.”

Try it

Take one vague prompt you used recently and rewrite it with goal, audience, constraints, and output format.

Myths

⚠️ Myth: Perfect prompts remove hallucinations.
✓ Reality: They help; verification still required.
⚠️ Myth: Longer prompts are always better.
✓ Reality: Noise and contradictions confuse models too.

Sources