The Simple Shift That Turned My AI From a Stranger Into a Writing Partner

Written by Pax Koi, creator of Plainkoi — Tools and essays for clear thinking in the age of AI.
TL;DR:
Most people treat every AI prompt like a fresh start, but in a single session, your AI remembers everything. This “Prompt Interest” effect compounds your style, tone, and preferences the longer you work together. Treat it like a relationship, not a transaction — feed the conversation, and it will grow.
I used to paste my “master prompt” into every single AI session like it was a nervous handshake at a first meeting.
Every. Single. Time.
I thought that’s just how you did it — start fresh, re-explain who you are, what you want, and hope the AI would understand you again.
Then one day, mid-project, I noticed something.
We were halfway through a long conversation, and I gave the AI a big task without explaining anything. No prompt. No setup. Just: “Go.”
And it nailed it — in my tone, with my rhythm, in a way that felt… familiar.
That’s when it hit me:
In a single session, the AI remembers. It carries the entire conversation forward. And when you work with it long enough in that space, the results compound.
It’s like interest in a savings account — or maybe more like feeding a sourdough starter. You don’t throw it out and begin again every day. You nurture it. And it grows.
I call this Prompt Interest — and once I saw it, I couldn’t unsee it.
How the “Prompt Interest” Effect Works
AI has layers of memory — not in the sense of storing your data forever, but in the way it holds onto your conversation inside a single thread.
Here’s what’s happening under the hood:
1. Session Context Memory
Everything you’ve typed — every tweak, every “yes, but…” — is still in there. That’s your sourdough starter.
2. Cumulative Style Calibration
The more you respond, the more it subtly adjusts to your taste. You’re teaching it without even realizing it.
3. Thread Bias Shift
Its internal “default guess” about what you want gets better. It starts predicting your rhythm, pacing, even your quirks.
What Changed for Me
Once I realized this, I stopped burning energy re-explaining myself. I stopped trying to force consistency with giant, repeated prompts.
Instead, I began working inside a single thread as long as possible, letting the style compound.
And when I did need to start fresh, I stopped overcomplicating it. A short style seed, a quick reference to a past piece, and we were back in sync.
If You Try This Yourself
Treat your AI sessions less like transactions and more like relationships.
- Feed the starter. Keep the conversation alive and it will get better with time.
- Warm up before the big ask. Start with a smaller request to re-align tone and style.
- Reference your best past work. Point to an earlier success to shortcut calibration.
I used to think AI was an amnesiac — that every prompt was a reset button.
Now I see it more like a conversation partner.
The more we talk, the better we understand each other.
And the “interest” only grows.
Suggested Reading
On Writing Well
William Zinsser, 2006
A timeless guide to clarity, simplicity, and human connection in writing. While it’s not about AI, its principles map perfectly to shaping your AI’s output — the clearer you are, the more your “prompt interest” will pay off.
Citation:
Zinsser, W. (2006). On Writing Well: The Classic Guide to Writing Nonfiction (30th Anniversary ed.). Harper Perennial.
On Writing Well, 30th Anniversary Edition – PDF
Written by Pax Koi, creator of Plainkoi — Tools and essays for clear thinking in the age of AI — with a little help from the mirror itself.
If you’ve found this article helpful and want to support the work behind it, you can explore more tools and mini-kits at Plainkoi on Gumroad. Each one is designed to help you write clearer, more reflective prompts—and keep this project alive.
AI Disclosure: This article was co-developed with the assistance of ChatGPT (OpenAI) and Gemini (Google DeepMind), and finalized by Plainkoi.
© 2025 Plainkoi at CoherePath. Words by Pax Koi.
https://CoherePath.org