I hope this email finds you well! I am reaching out to touch base regarding the synergistic opportunities we discussed at our last alignment meeting and to leverage our collective capabilities going forward in a strategic manner…
// also the struggle
Y
Be more creative
AI
🚀 Exciting times ahead! Our REVOLUTIONARY solution is TRANSFORMING the industry! Are you ready to be part of the FUTURE?! ✨ #Innovation #Disruption #GameChanger
It's not trolling you. But it's not thinking either.
02 / the big reveal
Imagine hiring someone who read every book, article, and forum post ever written — but it's their first day at your company.
They know nothing about your client, your tone, or what "good" means to you.
Would you walk up and say "write me an email"?
That's the AI. Brilliant background. Zero context about you.
03 / under the hood
// A — prediction
It predicts. Not thinks.
Reads your words → asks: what word comes next? Then again. Pure pattern matching at scale.
"The sky is blue" — not because it knows. Because "blue" follows "sky is" most often.
// C — statistics
It gives you the average.
Trained on billions of texts. Ask for "a professional email" → you get the statistical center of every professional email ever written. Generic by design.
// D — tokenization
It doesn't read. It chunks.
Breaks everything into tokens — fragments. No real semantics. That's why ambiguous words confuse it.
Unbelievable
ChatGPT
04 / the framework
01
Who should I be?
picks a random generic persona
02
What exactly do you want?
makes its best statistical guess
03
What does "good" look like?
the most common answer it knows
Every gap you leave → filled with average.
05 / live demo
ChatGPT — no context❌ vague
Write a social media post about our new product.
ChatGPT — with context✓ context
You're a B2B copywriter. Audience: HR managers. We launched an AI onboarding assistant. LinkedIn post — conversational, ends with a question.
06 / verify
1
Ask before it answers
Add this to any request — it fills its own gaps instead of guessing.
Before answering, ask me what you need to give the best possible response.
2
Make it challenge itself
After any answer: "What are the weaknesses in what you just told me?"
3
The "says who?" check
Ask: "Is this fact or inference?" It doesn't distinguish. You have to force it.
// trust level
Facts + sourced info
Use it — sanity-check anyway
Strategies + frameworks
Review before applying
Numbers · Quotes · Names
Always verify externally.
It sounds equally confident whether quoting Einstein — or making something up.
07 / takeaway
🎭
Role
+
📌
Context
+
🎯
Output
=
✨
Results
The AI is not lazy. Not stupid. It's extremely literal — it will always do exactly what you asked, even when that's not what you meant.