I used to think prompt engineering was just… typing smarter. bro I was so wrong.
Okay so listen, I gotta tell you something embarrassing. When I first heard the term AI prompt engineering, I literally thought it just meant “asking the AI in a fancy way.” You know, like adding please or using more words or whatever.
But no. That’s not even 5% of it.
What I’ve realized — after days of messing around, getting weird answers from ChatGPT, doubting my life choices, rage-closing tabs — is that prompt engineering is kinda like… therapy. Not for you. For the AI. You have to guide it, set the mood, give it a purpose, and sometimes even babysit it if you want good results.
It’s not just about what you type. It’s about how you say it. And who you’re telling it to be.
I used to just throw random questions at ChatGPT and hope for the best
Like literally, I’d be like:
“Hey write a blog post on freelancing.”
and it would respond like:
“Freelancing is a method of independent work…”
BRO. That’s not what I asked. That’s not what I meant. But also… I didn’t say what I meant. So yeah, my bad.
Here’s the hard truth:
AI’s not dumb. You’re just not clear.
(I had to tell myself that like 7 times.)
the prompt sandwich that saved my life (okay not literally but yeah kinda)
After crying internally for a week and reading 19 Reddit threads, I came up with this thing I call the Prompt Sandwich™. It sounds dumb, but it works.
So the structure goes like this:
👉 Top layer: context
Tell the AI who it is. Don’t just say “write something.” Say “You’re a sarcastic book critic from New York,” or “You’re a chill YouTuber who gives life advice.”
👉 Middle layer: your ask
Now tell it what to do. Not “tell me about X,” but “Write me a 200-word blog post explaining X in a funny, casual tone.”
👉 Bottom layer: output style
Want bullet points? A tweet thread? A TikTok script? Tell it. Format matters. If you don’t say it, it’ll give you boring paragraphs that sound like your 8th grade science textbook.
So yeah — role + task + style = ✨ usable results ✨
this one shift made my prompts 10x better overnight
Okay, I’ll keep this simple:
stop treating the AI like Google.
It’s not a search engine. It’s not a genie either. It’s more like a really fast intern who’s kinda dumb unless you give them super clear instructions. But once they get it? They kill it.
Like, instead of typing:
“How do I start a podcast?”
Try this:
“You’re a podcast coach who’s helped 100 people launch their first show. Write a short, casual list of 5 beginner tips someone can follow if they wanna start a podcast in 2024. Keep it friendly, no jargon.”
Now we’re talking.
mistakes I made (so you don’t have to)
This is me calling myself out:
❌ I used to write prompts like “write me something cool.”
(And then got mad when it was boring af.)
❌ I forgot to set a tone.
Sometimes I wanted humor but didn’t say that, so I got robot-corporate vibes.
❌ I overloaded the prompt with 10 instructions in one line.
Like “make it casual but professional but short but also detailed.”
What the hell does that even mean?? I don’t know. Neither does the AI.
❌ I never asked it to revise.
I’d take the first draft and be like “ugh this sucks” instead of just saying “rewrite this, but make it punchier.”
real tips from someone who’s still figuring it out (but kinda gets it now)
Here’s the stuff I actually use:
✅ Tell it who it is.
Literally start with “You are a…” and give it a job or personality.
✅ Use examples.
If you want something in a certain tone, paste a sentence and say “write like this.”
✅ Use step-by-step instructions.
Like:
- First, brainstorm 3 titles
- Then write a short outline
- Then write the actual post
It follows better when you chunk it out.
✅ Talk to it like a human.
You can literally say “That was close, but try again with more jokes.” And it’ll listen.
✅ Keep a doc of prompts that work.
Seriously, make a messy Notion page or a Google doc. Reuse and tweak. Saves so much brainpower.
okay but… why should you care?
Because this AI stuff isn’t going anywhere. Whether you’re a content creator, developer, student, freelancer, or someone just tryna survive Monday — knowing how to talk to AI is becoming a cheat code.
And not in some “become a billionaire overnight” way. But in a “save 3 hours of brain fog” kind of way. In a “holy crap this saved me from overthinking my email” kind of way.
Once you get how prompt engineering works, it’s like unlocking a second brain that helps you write, plan, brainstorm, fix your grammar, and roast your ex — all at once.
anyway. I’m outta chai. but you should try this.
Here’s what I want you to do.
Next time you open ChatGPT (or Claude or Gemini or whatever your flavor is), don’t just type the first thing that comes to your head. Pause. Breathe. Set the stage. Guide the vibes.
Say who it is. Say what you want. Say how you want it.
That’s it.
Do that, and your results will go from “meh, okay” to “yo this is scary good” REAL quick.
Alright I’m done. My Spotify just looped back to lo-fi again. Brain’s slowing down. Hope this helped.
Peace,
— Jexo (aka your slightly chaotic AI homie)
P.S. Save this blog. You will forget all this tomorrow when your brain is running on 2% battery. Trust me.
