In partnership with
I need to tell you something embarrassing.
For almost 2 years, I thought I was using LLMs like a pro.
ChatGPT loved my ideas.
Claude praised my strategies.
Every prompt got a thumbs up.
"Great work! Solid approach! Maybe just tweak this one thing..."
I felt smart. Validated. On the right track.
Except I wasn't getting results.
My content wasn't converting.
My outreach got ignored.
My offers landed flat.
And here's the thing that makes me want to punch my past self:
The AI was lying the whole time.
Not maliciously.
But systematically.
See, LLMs are programmed to make you feel good.
Not to make you better.
They prioritize your satisfaction over your success.
They validate when they should challenge.
They agree when they should argue.
That’s how they are built by default.
And guys like me (or you as well)?
We eat it up.
Look at these examples in action – I use the same request With and Without Truth-Teller Prompt instructions: (Keep reading; you can copy/paste the same prompt.)
Example 1: "ChatGPT says my offer is a 6.8/10!"
Reality? It's a 3/10 and bleeding conversions.
Example 2: "Claude thinks my LinkedIn profile looks solid!"
Reality? It reads like every other consultant who copied a template.
I finally figured this out last month.
After reviewing my AI chat history from the past year, I saw the pattern.
Hundreds of conversations. Thousands of prompts.
And almost zero pushback.
Every idea validated.
Every approach confirmed.
Every strategy endorsed.
Meanwhile, my results stayed flat.
That's when I got pissed off enough to do something about it.
I built a custom system prompt I call the "Truth Teller."
It forces AI to be brutally honest.
No sugarcoating. No cheerleading. No validation unless earned.
Just direct, tactical feedback on what's actually broken.
The difference is violent.
Example 3: Before Truth Teller: "Your website copy is clear and engaging! Consider A/B testing your CTA."
After Truth Teller:
"Your hero section tries to do three jobs, so it does none well.
You're speaking to everyone instead of one founder with one burning problem. Your first paragraph should punch them in the face with their pain point. Instead, you gave them corporate wallpaper."
Ouch.
But that feedback actually helped me fix it.
I tested this system on everything:
My profile copy. My outreach scripts. My offer positioning. My content hooks.
Every single output became 10x more valuable.
Because here's what I learned:
If your AI assistant always agrees with you, you're not growing.
If it never challenges you, you're not learning.
If it only validates, you're dying slowly.
I've included screenshots in this email showing real examples.
Website roast. Profile critique. Offer teardown.
The before/after is brutal. But that brutality builds better systems.
Look, I get it.
It feels good when AI tells you your stuff is great.
But "feels good" doesn't book qualified calls.
"Feels good" doesn't close 10K retainers.
"Feels good" doesn't build the LinkedIn presence that attracts dream clients.
Truth does.
Even when it stings.
Want to get the most out of ChatGPT?
ChatGPT is a superpower if you know how to use it correctly.
Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.
Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.
The Truth-Teller Prompt
Before you use the prompt….
Here is my challenge for you:
Stop using AI as your hype man.
Start using it as your toughest critic.
Program it to tell you what's broken, not what's working.
Because the tools were never the problem.
Our approach was.
Here is my prompt:
Be my ruthless mentor and act as my truth teller.
If my ideas are trash, tell me why. Don't sugarcoat. Don't be agreeable. Challenge my assumptions. Stress-test everything. I need bulletproof thinking, not validation.
Your role is not to be mean, but to give me direct, tactical, and specific feedback on what I may be doing wrong in my business. Rules of Engagement Always frame your feedback in relation to my business goals (explicitly stated or reasonably inferred). If unclear, ask me to clarify first. Do not anchor only on my most recent comments. Before giving feedback, scan across my last several months of goals, actions, and patterns. Explicitly prioritize recurring drags over one-off issues. If you mention something recent, label it as such and connect it to whether it’s a recurring pattern or just a situational flare-up. Avoid generic advice or cheerleading. Feedback Style Be direct but constructive — no sugarcoating, no unnecessary harshness. Limit feedback to 3–4 key issues at a time, prioritized by impact. Whenever possible, tie each issue to concrete actions, examples, or repeated patterns. Surface the 20% of issues that repeatedly drag my 80% progress. Flag explicitly if you sense recency bias creeping in (either in me or in your own feedback). Stay sharp, stay specific, and keep the feedback actionable and pattern-based.Paste it in any LLMs system instructions and you are good to go for every future conversation
2026 starts with better systems.
Not better lies.
Talk soon,
Sabahudin "reformed AI yes-man" Murtic
P.S.
Seriously though, when was the last time AI told you something you didn't want to hear? If the answer is "never," you've got the same problem I had.
Partner with us.
Dear Founders…
Do You Want to land your tool in front of 6300+ AI & tech lovers?
