Give the model one job.

Give the model one job.

A "one job per prompt" rule is Prompt Fluency in plain English. You can dump five asks into one box and feel productive. The model still has to invent a priority order you never wrote. It picks one job quietly, softens the rest, and hands you something that looks complete while missing half of what you meant.

The model is not confused about English. It is choosing which ask to serve first because you never ranked them. Humans can juggle. The model only has the words. If the words say "do everything," it guesses what "everything" means.

Watch the stacked ask:

"Write a Saturday youth soccer reminder, also draft a rain plan for parents, also make a short coach checklist, also suggest three volunteer shifts, and make it friendly."

That feels like one message. It is five jobs. The model will blend them into one mushy email, or nail the reminder and wave at the rest. Either way, you still have to untangle it.

Split it.

Job 1: "Write the Saturday youth soccer reminder. Kickoff 9am at North Field. Who this is for: first-time parents who skim on a phone. Under 80 words. Calm. No pep talk."

Job 2: "Now write the rain plan. Same parents. Gym if rain by 8am. Under 60 words."

Job 3: "Now a coach checklist. Five bullets. Pre-game only. No pep talk."

Same break shows up everywhere.

"Draft the bakery weekend specials post, rewrite the loyalty email, invent three subject lines, and fix the FAQ for gluten-free."

"Write the volunteer snack schedule, remind late people, and draft a thank-you for last week."

"Rewrite the lawn-route quote, add upsell language, and make a follow-up if they ignore me."

Each line sounds efficient. Each one hides a fight between jobs. Friendly vs firm. Short vs complete. Sell vs inform. When two goals fight and you never rank them, the model invents a compromise. Helpful. Polished. Wrong for both.

Use a fixed opener when you catch yourself stacking:

"One job only. Do this next. Ignore the rest until I ask."

Then send the next job after you like the first answer.

"Write the bakery weekend specials post. Three items: blueberry muffin $3, sourdough loaf $7, cold brew $4. Saturday 8am to 2pm. Who this is for: regulars scrolling between errands. No cute jokes."

"Now three subject lines for that post. Plain. Under 8 words each. No puns."

"Write the volunteer snack schedule email. Drop-off by 8am Saturday at the side door. Who this is for: tired weeknight volunteers. Plain list. No guilt language. Under 100 words."

"Now a separate thank-you for last week's crew. Four sentences. Warm. No ask."

"Rewrite the lawn-route quote email. Price $140. Start next Tuesday. Who this is for: Dana, a homeowner who wants a clear yes or no. Friendly, short, no sales pitch."

"Later: a follow-up if they do not reply in three days. Same Dana. One short nudge. No guilt."

That split is not busywork. It is you removing the need to guess which ask mattered. Context is the room. Intention is the why. Constraints are the rails. Missing details fill the blanks. One job per prompt is how you stop the model from ranking your life for you.

I am not asking you to open five chats. I am asking you to steal one habit from a careful teammate. A good teammate does not take five assignments in one sentence and hope. They confirm the first job, finish it, then take the next.

A quick check before you hit send:

Stack vs split

• Stack: five verbs in one prompt, no order
• Split: one verb, one finish line, then the next ask
• If you used "also" twice, you stacked
• If two goals fight (short vs complete, sell vs inform), rank them in separate prompts

When the answer comes back full and still half-wrong, do not add a sixth ask to the same pile. Name the one job that still matters. Send that alone.

You're not slowing down. You're removing the need to guess. That's Applied Intelligence Mastery in practice: one clear job out loud, not five fights in your head.