Introduction
Trust me — as a non-native English speaker, even if you can read technical documentation fluently (or mostly fluently), speaking and writing are entirely different beasts. In the following sections, I’ll show you some examples where I provided rough translations or drafts — and how ChatGPT helped inject personality into topics that are otherwise dry, robotic, or just plain boring.
AI prompt guiding
Guiding the AI: Defining Roles and Behavior One of the most important lessons: you need to tell ChatGPT exactly who it is, how it should behave, and what role it plays in the context of your prompt. You can shape its responses to be:
- Fact-driven and concise
- Creative and opinionated
- Polite or informal
- Serious, humorous — or somewhere delightfully in between
As long as you stay within the same chat, it will remember the context — including the tone, personality, and previous discussions. This makes it easier to maintain flow and consistency throughout longer tasks like documentation or blog writing.
AI - If you can’t convince them, confuse them
While this is a perfect strategy for meetings, don’t do it on the prompt.
Personally, I try to keep chats clean and avoid jumping wildly between topics. If something new, unclear, or unrelated pops up — I’ll often open a separate chat, even anonymously, just to avoid polluting the context we’ve already built. It’s a small habit, but it helps keep the output sharp and focused — especially when working on a complex project like this one.
Defining the role clearly (“You are a native English-speaking co-writer with a sense of humor…”):
This really helps steer tone and output — it sets the personality baseline.
Being explicit about behavior and output format (“Be factual, allow humor, stay polite…”):
The more specific you are, the more consistent and relevant the responses.
Staying within a focused thread for a task preserves memory and context so that the assistant doesn’t start contradicting or derailing itself.
Opening new chats for unrelated or exploratory topics is a pro-level habit. It prevents context pollution, which is crucial when you’re crafting something structured — like a long-form blog, doc series, or detailed technical breakdown.
What Might Change (or Be Worth Knowing)
Session context isn’t infinite: Over time (in a long chat), some earlier parts may fade. If you’re working on a large document across multiple sessions, it’s smart to recap your setup or paste key context back in now and then. Instructions can be reinforced: If the AI starts drifting in tone or verbosity, a gentle reminder like “Stick to the dry-but-humorous voice, please” helps reset things.
Prompt engineering is evolving but we need to master basics — clarity, structure, tone-setting — will stay relevant even as models grow more powerful.
Translation samples
Sample 1 (introductory text)
Let’s look at the introductory text above. My original text read:
Trust me, as a non native english speaking person, even if you can read technical docs more or less fluently, speaking and writing is a different story. In the following sections I will show you some translations and how ChatGPT added personality to the “dry and boring” topics.
It is important to exactly tell ChatGPT who he is, how he should behave and which role he has in the context of the prompt. You have to define how he responds, from factual driven to generate whatever you want, stay polite, allow humor aso. As long you use the same chat he will remember the context from personality to previous discussions. I try to keep chats clean and do not jump between topics. In case an interesting or unclear part enters the stage I tend to immediately open a new chat (even anonymous) to not water down what we have achieved by adding foreign context to the current discussion
Sample 2 (Overview)
Let’s look at the overview text. My original text read:
The Beginning (a.k.a. The Five-Year Think-Tank) This project started — conceptually — about five years ago. For a long time, it lived entirely in my head, evolving slowly, quietly, and occasionally waking me up at 3 a.m. with brilliant ideas and equally terrible ones.
Now, in May 2025, it’s finally taking shape in the real world. And oddly enough, it all began by finding a blogging platform that didn’t make me want to rage-quit. Without that, I couldn’t share any of this with you.
Over the next few months, I’ll show how modern AI — yes, tools like ChatGPT — can accelerate progress, even when you’re diving into a completely new programming language or toolchain. You’ll see how I use it as a kind of rubber duck with a PhD and infinite patience.