"Make it better" is not a prompt.

A "make it better" prompt feels like a follow-up. You already got a draft. You know it is not right. So you type three words and hit send. Improve this. Fix this. Make it pop. Then the second answer is still wrong, just in a different costume. The model did not ignore you. You never named the target.

Prompt Fluency is communicating so the model does not have to guess. "Better" is a blank. Better for who. Better than what. Better by which measure. Humans hear "better" and fill it from shared memory. The model only has the words. It fills the blank with something statistically reasonable. Cleaner prose. More adjectives. A friendlier tone. Rarely the fix you meant.

Try this exchange:

You: "Write a flyer for Saturday soccer camp."

Model: a generic, cheerful flyer.

You: "Make it better."

Now what. Shorter. Louder. More parental. Less salesy. Ages on the page. Rain plan in the footer. The model cannot know which "better" you meant, so it polishes the wrong surface.

Same break shows up at work.

"Rewrite this volunteer email. Make it better."

"Improve this lawn-route quote letter."

"Fix this bookstore product blurb."

The ask sounds like direction. It is not. Direction would name the miss.

Say the miss out loud, then name the finish line.

"Rewrite the soccer camp flyer. Parents skim on their phone. Put ages 7 to 9, drop-off at 8:30am, and the pavilion rain plan above the fold. Cut the pep talk. One page."

"Rewrite the volunteer snack email. Keep the March rotation. Call out that the Rivera family is out the second week. Warm, not cute. Under 120 words."

"Rewrite the lawn-route quote. Eighteen yards, about 40 minutes each. Plain monthly total. No upsell language. Assumptions listed at the bottom."

"Rewrite the Harbor Books blurb for the hardcover only. Separate print from ebook. No spoiler. Under 80 words."

That is not "better." That is a job. Context, intention, constraints, and the boring details are doing the work. The model can finally aim.

I am not saying you need a perfect first prompt. First drafts miss. Fine. The follow-up is still a prompt. Treat it like one. Name what failed. Name what wins. If you cannot say those two sentences, the model will invent a nicer version of the same miss.

A quick check before you type "make it better":

Vague vs aimed

• Vague: improve, fix, polish, punch up, make it better
• Aimed: what is wrong, what "done" looks like, who reads it, what must stay or go
• If the follow-up only has the vague word, you have not prompted yet

When the second answer comes back polished and still wrong, do not keep stacking soft adjectives. Rewrite the ask. Point at the miss. Put the rails in. Type the obvious facts. Then send again.

You're not asking for vibes. You're removing the need to guess. That's Applied Intelligence Mastery in practice: clear communication on the second send, not just the first.