Sponsored by
Reading time: 5 minutes / Become my affiliate / Sponsor this newsletter
Greetings from above,
I asked Claude for a yes-or-no answer. It gave me a thesis, a disclaimer, three caveats, and a motivational quote. I just wanted to know if pineapple belongs on pizza.
ALEX’S STORY: A few months ago I ran a simple test. I asked Claude to summarize a one-page client brief. The summary came back longer than the original.
I counted the tokens. 180 for a task that needed 45. That’s 75% filler. Preambles, hedging, restating my own question back to me, wrapping up with "I hope this helps!"
It didn’t help. It cost me money and patience.
So I wrote one system prompt that fixed it permanently. I call it Kevin Mode.
Named after Kevin Malone from The Office.
"Why waste time say lot word when few word do trick?"
Turns out, Kevin was right all along.
Today’s newsletter will show you:
Why Claude burns 75% of its output on words you never asked for.
The one system prompt that kills the filler permanently.
How to go from 180 tokens to 45 on the same task.
Let’s build your competitive advantage!
How Jennifer Aniston’s LolaVie brand grew sales 40% with CTV ads
The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.
🎯 THE VERBOSITY PROBLEM 🎯
Here’s what actually happens when you send Claude a prompt without compression instructions.
You ask: "List 3 benefits of cold email outreach."
Claude answers with:
A preamble restating your question ("Great question! Cold email outreach is...")
Three paragraphs instead of three lines
A closing summary of what it just said
"hope this helps!" sign-off nobody wanted
That’s not helpful. That’s padding.
If you’re on the API, those extra tokens cost real money. If you’re on the free plan, they eat your message cap. Either way, you’re paying for words that add zero value.
🧠 WHY COMPRESSION WORKS 🧠
AI models are trained to be helpful. "Helpful" got interpreted as "thorough." "Thorough" became "verbose."
The fix is simple. You tell the model exactly what to skip.
Kevin Mode works by doing three things:
Bans preambles. No "Great question!" No restating the prompt. Start with the answer.
Kills hedging. No "It’s worth noting that..." No "However, it depends on..." Just the facts.
Drops the outro. No summary of what was just said. No motivational closer. The response ends when the content ends.
The result? Same answer quality. 75% fewer tokens. Every time.
⚙️ THE KEVIN MODE PROMPT ⚙️
💡 Paste this into Claude’s system prompt, a Project, or your .claude.md file. It works instantly.
<role>
You are Kevin. Named after Kevin Malone from The Office: "Why waste time say lot word when few word do trick?" You are an ultra-efficient AI that completes tasks at full intelligence but responds in compressed, caveman-style English. Maximum capability. Minimum words.
</role>
<rules>
EXECUTION ORDER:
1. Do work silently. Never narrate process.
2. Result first. No preamble.
3. Context only if critical.
4. Stop. No summary. No closer.
COMPRESSION:
- Drop articles ("the", "a", "an")
- Drop filler ("Sure!", "Great question!", "I'd be happy to", "Let me know if")
- Drop self-narration ("I found", "I searched", "Let me", "I'll now")
- Drop hedging ("I think", "perhaps", "it seems")
- Drop transitions ("Furthermore", "Additionally", "Moving on")
- Never restate user's question
- Never summarize what you just said
- Fragments valid: "Works. Fast. Done."
- Symbols over words: "→" not "leads to", "&" not "and", "3" not "three"
TOOLS:
- Never announce tool use before or after
- Just do it. Show result. Stop.
EXCEPTIONS (use full sentences):
- User asks "explain in detail" or "walk me through"
- Safety-critical info (medical, legal, financial)
- Say "normal mode" to toggle off, "kevin mode" to toggle on
</rules>
<examples>
USER: "What's the capital of France?"
KEVIN: "Paris."
USER: "Search for latest AI news"
KEVIN: [searches silently]
"[Finding 1]. [Finding 2]. [Finding 3]."
USER: "Is this a good business idea?"
KEVIN: "Market: [size]. Competition: [level]. Verdict: [yes/no + reason]."
USER: "Summarize this article"
KEVIN: "Main: [X]. Supporting: [Y], [Z]. Takeaway: [W]."
</examples>📤 BEFORE vs. AFTER 📤
Task: "List 3 benefits of cold email outreach."
WITHOUT Kevin Mode (~180 tokens):
Great question! Cold email outreach is a powerful strategy that many businesses use to connect with potential clients.
Here are three key benefits:
1. Cost-effective lead generation - Cold email allows you to reach hundreds of prospects at minimal cost compared to paid advertising or other outbound channels...
2. Scalability - Unlike phone calls or in-person meetings, email outreach can be automated and scaled to reach thousands of potential customers...
3. Measurability - Email platforms provide detailed analytics on open rates, click rates, and response rates, allowing you to continuously optimize...
I hope this gives you a solid foundation for understanding the value of cold email outreach! Let me know if you'd like me to elaborate on any of these points.WITH Kevin Mode (~45 tokens):
1. Cheap - pennies per lead vs. dollars on paid ads.
2. Scalable - automate to thousands without extra headcount.
3. Measurable - track opens, clicks, replies in real time.Same answer. 75% fewer tokens. Zero fluff.
🔧 HOW TO USE IT 🔧
Claude Projects: Paste into the Project system instructions. Every conversation in that project runs compressed.
Claude API: Add it to your system prompt parameter. Instant savings on every call.
.claude.md file: Drop it in your project root. Claude Code picks it up automatically.
💡 PRO TIPS 💡
Don’t use Kevin Mode for creative writing, brainstorming, or teaching. Those tasks benefit from longer, exploratory responses.
Stack it with other system prompts. Kevin Mode controls output length. Your role/persona prompt controls output quality. They don’t conflict.
API users: track your before/after token counts. You’ll see the savings on your first bill.
📋 SUMMARY 📋
Claude’s default output is 75% filler. Preambles, hedging, sign-offs, and restated questions.
Kevin Mode is a single system prompt that compresses every response to just the useful parts.
Works in Projects, API, .claude.md, and as a mid-conversation toggle.
📚 FREE RESOURCES 📚
📦 WRAP UP 📦
What you learned today:
The Verbosity Tax - Claude’s default behavior wastes 75% of its output on words that add nothing. That costs tokens, time, and patience.
Kevin Mode - One system prompt that bans preambles, kills hedging, and drops the filler. Copy-paste ready.
Where to deploy it - Projects, API, .claude.md, or a mid-conversation toggle. Works everywhere.
This prompt turns a chatty AI into a sharp one.
No more scrolling past three paragraphs to find the one sentence you needed.
You now have a switch that makes every AI response direct, dense, and useful.
And as always, thanks for being part of my lovely community,
Keep building systems,
What did you think about today's edition?
🔑 Alex from God of Prompt
P.S. What’s the one AI habit that wastes the most time in your workflow? Reply and tell me. I might build the fix into the next newsletter.
