The model can only work with what you give it
A large language model has read more text than any person ever will, but it knows nothing about your request beyond the words you send. When an answer disappoints, the cause is usually a prompt that left the model guessing: what the task really was, who it was for, or what a good answer looks like.
Prompt engineering is the practice of writing prompts that leave nothing important to guess. It sounds technical, but it's closer to writing a good brief for a colleague than to programming.
Why it matters
- Fewer rewrites. A clear prompt gets a usable answer on the first or second try, not the fifth.
- Lower cost. Every retry spends time and, in Ace, credits. Getting it right sooner is cheaper.
- Repeatable results. A good prompt can be saved and reused, so the tenth answer is as good as the first.
- Portable skill. The techniques in this course work with Gemini, Claude, GPT, and whatever comes next.
What a better prompt looks like
Vague: "Write about time management."
Clear: "Write a 150-word LinkedIn post for freelance designers who miss deadlines because they take on too much. Give one practical rule they can use this week. Friendly, direct, no buzzwords. End with a question."
The second prompt isn't longer for the sake of it. Each added phrase answers a question the model would otherwise answer for you, usually with the most generic option.
Prompts are conversations, not commands
You rarely get the perfect answer in one message, and that's fine. A good first prompt gets you close, and a good follow-up ("shorter, and lead with the rule") gets you the rest of the way. Module 4 covers multi-turn design in depth.
Try it
Open Ask Ace and send a request you'd really make this week, written the way you'd normally write it. Save the prompt and the answer: they're your "before" sample, and you'll rewrite that prompt in the next lesson.