ð This time, we will cover the foundation of the series: prompt engineering. I will explain the basics of how to refine a single instruction and why saying ‘You are an expert in…’ is no longer enough, using examples from English learning.
Introduction
ð Erica: ‘For those of you thinking, “Weren’t we going to talk about loops? Why prompt engineering now?” don’t worry. Prompts are the foundation upon which context, harnesses, and loops are built. If this part is vague, everything that follows will be shaky, so let’s take a good look at it.’
Last time, I talked about how our relationship with AI has evolved through four stages: Prompt -> Context -> Harness -> Loop. This time, we are looking at the very first stage: prompt engineering.
ð What is Prompt Engineering?
ð¹ The stage of refining ‘how to ask’ AI
Prompt engineering is the act of refining the individual instructions sent to an AI. Even when asking for the same content, the quality of the response changes significantly depending on how you ask.
ð¹ The basic form in English learning
When using AI for English learning, you are likely already refining how you ask on a daily basis. Asking ‘Correct this English sentence’ is less effective than ‘Correct this English sentence to sound like a natural business email’ to get the answer you want. This is the simplest form of prompt engineering.
ð Why ‘You are an accounting expert’ is not enough
ð¹ The limitations of the role-assignment prompt mentioned last time
In the prompt set introduced in the paid section last time, there was an example like this:
‘You are an accounting expert. Please tell me three common mistakes in expense reporting.’
This role-assignment technique is often introduced as an introduction to prompt engineering. But actually, as it stands, it’s a bit outdated. Just giving a role doesn’t tell the AI ‘who’ wants the answer or ‘why’ they want it.
ð¹ What information does AI really need?
What is truly important to an AI is not the title, but the ‘purpose’ and the ‘goal.’ Why do you want to know this, and who will use the answer and how? If this is conveyed, you will get a sufficiently accurate answer without having to specify the role in detail.
ð Modern writing style â Adding purpose, goals, and conditions
ð¹ Conveying the purpose
Just adding a word about ‘why you want to know this’ changes the angle of the AI’s response. Adding just the phrase ‘because I am a beginner in accounting’ will cause it to lean toward an explanation that avoids technical jargon.
ð¹ Conveying the goal
“Who is using it and how” is just as important. If you add “in a format I can use immediately for work,” you will get closer to a practical answer rather than a textbook generalization.
ð¹ Communicating Conditions
For English learning, adding conditions like “for business emails” or “using natural, not-too-formal phrasing” is a type of this refinement. For current AI, this way of writing is much easier to understand than assigning a role.
Conclusion
ð Erica: “Next time, we’ll cover context engineering. We’ll move beyond just how to ask once and talk about ‘what to show the AI.’ From here on, there’s a set of Japanese-English prompts using today’s content, plus an exclusive, cute-and-sassy bromide image of Erica (vertical) cheering you on, so take a look. It’s sized to be a smartphone wallpaper, too.”
From here on is a set of “Japanese prompts x English prompts + exclusive Erica images” that you can use exactly as they are for this topic. As a way to practice English, please take a look at both prompts together.
â¼ Paid content starts here (200 yen) â¼
