The Human Prompting Playbook
How to work with AI without outsourcing your judgment, diluting your thinking, or losing your edge
This is not a list of magic prompts.
It’s not a toolkit for making AI sound smarter.
And it’s definitely not another “50 ways to prompt better.”
This is something else:
A way to think with AI without losing your internal compass.
A way to stay sharp, aligned, and honest while moving fast with systems that never slow down.
Whether you’re writing a blog post, shipping infra, debugging a pipeline, or shaping a strategy doc, the risk is the same:
Moving faster than you can feel
Letting AI fill in your thinking
Optimizing away the part of the work that actually matters: your judgment
This playbook is how I’ve kept that edge intact.
It’s how I wrote the most honest work of my life.
And it’s how I ship real code with AI, without handing over control.
Not for performance.
For alignment.
Here’s the shift that changed everything for me:
Most people use AI to finish things faster.
I use it to think more honestly.
That shift became the foundation for everything that came next.
These are the five practices I actually use, across writing, engineering, and everything else.
🧠 Five Real-World Practices That Actually Work
1. Start with friction not a task.
Old prompt mindset:
“Write me a blog post about AI.”
“Summarize this code.”
What I do instead:
Start with what doesn’t sit right.
“Every time I feel like AI is going to replace us, I open YouTube and watch someone talk about something they care about. I’m not sure what that means yet, can we unpack it?”
That line became the core of an entire piece.
But I didn’t ask for a headline.
I asked to examine tension.
Same in engineering:
“Here’s what’s happening when I run this GitLab step locally, it’s slower than expected, and something about the env variable behavior feels inconsistent. Can we think through what I’m missing?”
Friction is where insight begins.
Try this:
Before prompting, write down where you feel stuck, unsure, or unsettled.
Then ask:
“Can you help me see what I’m not naming yet?”
2. Build with the AI not through it.
Old prompt mindset:
“Generate a full plan.”
“Write the post.”
“Architect the service.”
What I do instead:
I treat AI like a second mind, not an answer machine.
“Let’s structure this section together, but don’t write it yet.”
“Here’s what I’m thinking about the Sonic dev workflow, push on this before we write the doc.”
This turns the conversation into a feedback loop, not a vending machine.
Try this:
“Let’s co-design this part first, help me challenge my assumptions.”
“Here’s what I believe so far. Where would this break?”
3. Surface what’s unstated not just what's next.
Old prompt mindset:
“Finish the paragraph.”
“What’s the next step?”
What I do instead:
Use the AI to check for what I haven’t said out loud.
In writing:
“I’ve drafted the section. What am I implying here but not admitting?”
“Where am I being safe instead of clear?”
In engineering or product:
“Here’s my current migration plan. What risk am I avoiding naming directly?”
“Is there a design tradeoff I’m bypassing by calling this MVP?”
This stops you from hiding behind velocity.
It makes invisible logic visible, fast.
Try this:
“Where am I skipping over a hard decision here?”
“What am I assuming that I haven’t validated?”
4. Reduce, don’t decorate.
Old prompt mindset:
“Make it sound more professional.”
“Add more details.”
What I do instead:
Ask the AI to strip it back to what I actually mean.
“Cut the fluff. What’s the real point?”
“Remove the architecture buzzwords, what is this system really doing?”
The more honest the language, the clearer the design.
This works in docs, team emails, release notes, and even Slack.
Try this:
“Rewrite this like I’d explain it to my teammate, no slides.”
“What’s the raw logic behind this, not the jargon?”
5. Write while you build. It’s not output it’s clarity.
Old prompt mindset:
“Once I’m done, I’ll write the summary.”
“I’ll document this after I ship it.”
What I do instead:
I write while I’m building, to think.
“Help me shape this README while I design the service.”
“Let’s write the ‘Why’ section now, before we lock the migration.”
This exposes weak thinking early.
It prevents rework.
It makes decisions sharper.
And when writing isn’t about polishing it becomes runtime, not packaging.
Try this:
“I’m midway through building, let’s write the doc now to surface what’s unclear.”
“Can we write the first paragraph before we start the slides, just to see what’s alive here?”
💭 The Prompt I Actually Use
When I’m not sure what to say, or how to begin, I start here:
“I’m trying to think clearly.
I want signal, not noise.
Let’s go one part at a time, and build this together.”
That’s it.
That’s what opens everything that matters.
🧭 Final Word
You don’t need more prompt templates.
You don’t need to engineer the perfect question.
You just need to:
Start from tension
Think with the system, not through it
Don’t hide what you’re skipping
Strip the language down
Write as you go
That’s not prompting.
That’s staying real while you build.
That’s how I work with AI.
Not for output.
For alignment.
Every time.