Prompt Engineering for Beginners (Part 4): Roles, Rules & the Power of Clear Writing

In Part 3, we explored six useful prompting techniques, including role-based and constraint-based prompting.

We learned that a role can guide how AI approaches a task, while constraints explain the rules it should follow.

But there's an interesting question:

What happens when we combine a clear role with specific writing rules?

Can we make AI explanations easier to understand without losing important information?

To explore this, we'll look at something unexpected: a writing standard originally developed for aircraft maintenance instructions.

It's called ASD-STE100, or Simplified Technical English.

Don't worry about the complicated name. The idea behind it is surprisingly simple.

1. Role-Based Prompting: More Than Just "Act as an Expert"

Imagine you want to learn about personal budgeting.

You ask AI:

Explain budgeting.

You might get a useful answer, but it could be too broad.

Now try:

Act as a friendly financial educator.

Explain budgeting to someone
who has never managed a budget.

Use simple English and
one everyday example.

The second prompt gives the AI a clearer direction.

It tells the model how to approach the explanation and who the intended reader is.

That's role-based prompting.

What makes a useful role?

A helpful role usually explains more than a job title.

Notice that you don't need to invent someone's experience or qualifications.

You could write:

You are a financial expert with 30 years of experience.

But that doesn't give the AI 30 years of real experience.

A clearer instruction would be:

Explain basic financial concepts carefully. Mention uncertainty and avoid unsupported advice.

The key lesson: A specific role can help guide the response, but it doesn't guarantee expertise or accuracy.

2. Constraint-Based Prompting: Set the Rules

Imagine you're organizing a school event.

You ask someone to prepare a speech, but forget to mention that the speech must be under two minutes.

They might write something excellent that you cannot use.

The problem is that an important requirement was missing.

The same thing happens with AI.

A constraint is simply a rule or limit that the answer should follow.

Without constraints

Write a speech about teamwork.

With constraints

Write a speech about teamwork.

Requirements:
- Suitable for school students
- Under 200 words
- Simple English
- Include one everyday example
- End with a positive message

Now the expectations are clearer.

Two types of constraints

A hard constraint is something you can check directly.

For example, you can count whether the answer exceeds 200 words.

A soft constraint is more subjective. Different readers might disagree about what sounds friendly.

Are more rules always better?

Not necessarily.

Too many instructions can make a prompt confusing, especially if they conflict.

For example:

Explain this complicated topic in detail using only ten words.

That's probably unrealistic.

Instead, focus on the rules that genuinely matter.

Good constraints make expectations clearer. They don't make the AI perfect.

3. Why "Write Clearly" Isn't Always Clear Enough

Here's a small experiment.

Ask AI:

Explain how a refrigerator works.

Write clearly.

What does clearly mean?

Does it mean short sentences?

Simple words?

An everyday example?

No technical terms?

Different readers may understand that instruction differently.

Now try:

Explain how a refrigerator works.

Rules:
- Use everyday English.
- Keep sentences short.
- Explain unfamiliar terms.
- Use one idea per sentence.
- Include a simple example.

This gives the model more specific guidance.

We have changed a vague request into instructions that are easier to follow and evaluate.

And this is where technical writing standards become interesting.

4. ASD-STE100: What Can Aircraft Manuals Teach Us?

Imagine you're an aircraft maintenance technician.

You're reading instructions about inspecting an aircraft component.

The instructions must be clear because misunderstanding an important step could have serious consequences.

Now imagine that technicians around the world speak different first languages.

Complicated sentences and unfamiliar words could make the instructions harder to understand.

To address this problem, the aviation industry developed a controlled form of English.

It became known as Simplified Technical English, or STE.

Today, ASD-STE100 provides formal rules and vocabulary for writing technical documentation.

The standard developed from work that began in the late 1970s. Its current Issue 9 was released in January 2025.

You can learn more on the official ASD-STE100 website.

What's special about it?

STE isn't just about using fewer words.

It aims to make instructions clearer and more consistent.

It uses two major elements:

1. Writing rules

These guide how sentences and instructions should be written.

2. A controlled dictionary

This identifies approved words and their permitted meanings.

The goal is to reduce unnecessary ambiguity.

Let's explore the ideas we can borrow for everyday AI prompting.

5. Five Writing Principles That Can Improve AI Responses

You don't need to study the complete ASD-STE100 standard to benefit from its approach.

Here are five useful principles inspired by clear technical writing.

Principle 1: Use shorter sentences

Before:

In order to successfully complete the registration process, users are required to provide the necessary information before proceeding.

After:

Enter the required information. Then complete your registration.

The second version is easier to read.

Principle 2: Use direct instructions

Before:

The form should be submitted by the user.

After:

Submit the form.

The second sentence tells the reader exactly what to do.

This illustrates the idea of active voice.

In simple terms, active voice makes it clearer who performs an action.

Principle 3: Use familiar words

Simple words often make information easier to understand.

However, some technical words are necessary. The goal is to explain them, not remove important meaning.

Principle 4: Keep terminology consistent

Suppose instructions use the words button, control, and switch to refer to the same thing.

The reader might wonder whether they mean different objects.

Using the same term consistently can reduce confusion.

Principle 5: Give one instruction at a time

Before:

Open the application, select your profile, update your information, and save your changes.

After:

1. Open the application.
2. Select your profile.
3. Update your information.
4. Save your changes.

The steps are easier to follow and check.

These are useful writing practices, but these examples are not claims of official ASD-STE100 compliance.

6. Can We Ask AI to Follow ASD-STE100?

Yes. We can ask AI to use writing principles inspired by the standard.

Andrej Karpathy, a well-known AI researcher, discussed using ASD-STE100-style instructions in October 2026.

He also mentioned asking for something like "80% of the way to ASD-STE100", because the complete standard is strict.

That's an interesting prompting idea.

But what does 80% mean?

It isn't an official certification or formal compliance level.

It's an informal way to ask AI to approximate the style without strictly implementing the entire standard.

A practical prompt

Act as a clear technical writing assistant.

Explain [TOPIC] to a beginner.

Use writing principles inspired
by Simplified Technical English.

Requirements:
- Use short, direct sentences.
- Prefer familiar words.
- Explain technical terms.
- Keep terminology consistent.
- Give one instruction at a time.
- Preserve important facts and numbers.

If information is unclear,
mention the uncertainty.

Do not claim official
ASD-STE100 compliance.

You can replace [TOPIC] with anything you want to learn.

For example:

• How an internet connection works
• How to use a mobile banking application
• How solar panels generate electricity
• How to reset a password

A small technical detail

The formal standard includes specific writing requirements and approved vocabulary.

For example, it uses different sentence-length limits for procedural and descriptive writing, including the familiar 20-word and 25-word limits.

But simply telling AI to follow those numbers does not prove that the result satisfies the full standard.

Writing in an STE-inspired style and meeting official ASD-STE100 requirements are different things.

7. Let's Compare Two Explanations

Suppose we want AI to explain cloud storage.

Basic prompt

Explain cloud storage.

An AI might generate something like:

Cloud storage is a digital data management solution that enables users to store information on remote servers accessed through internet-based infrastructure, providing scalability, accessibility, and synchronization across multiple devices.

This explanation isn't necessarily wrong.

But someone unfamiliar with technology might struggle with terms like infrastructure, scalability, and synchronization.

Clear-writing prompt

Explain cloud storage to someone
with no technical background.

Use short sentences and
simple everyday words.

Explain one idea at a time.

Include a familiar example.

A possible response:

Cloud storage lets you save files on computers connected to the internet. You can open those files from different devices.

For example, you can upload photos from your phone to an online storage service. Later, you can open those photos on your laptop.

You usually need an internet connection to upload files or access files that are not already saved on your device.

See the difference?

The second explanation uses shorter sentences and a relatable example.

Important: These are illustrative responses, not measured results from an actual model comparison.

To find out whether a prompt improves responses consistently, we would need to test it.

8. Why Clear Writing Doesn't Always Mean Correct Writing

Here's something important.

Imagine an AI produces this sentence:

The Earth has two moons.

It's short.

It's simple.

It's easy to understand.

But it's incorrect.

This shows that clarity and factual accuracy are two different things.

A writing standard can help make information easier to read.

It cannot automatically guarantee that the information is true.

The same applies when we ask AI to follow ASD-STE100 principles.

Even if the response looks clear and professional, it may still contain incorrect facts or misuse approved terminology.

What about official compliance?

Official ASD-STE100 compliance requires checking the actual rules and controlled dictionary.

For example, a word might be permitted in one meaning but not another.

An AI could use a familiar word in a way that sounds natural while still violating the standard.

For professional technical or safety-related documentation, proper verification remains necessary.

9. For the Curious: How Would Professionals Check Compliance?

This section is optional for readers who want a little more technical detail.

If a company needs documentation that genuinely follows ASD-STE100, simply asking AI to write in that style isn't sufficient.

A more reliable process could look like this:

This may involve specialized software and qualified reviewers.

It also connects with a broader prompt-engineering principle:

When exact requirements matter, don't rely only on AI instructions. Verify the final output.

You don't need this level of checking for an everyday email or a school explanation.

But it becomes important when documentation must meet a formal standard.

10. Combining Roles and Constraints in Everyday Tasks

Let's put the concepts together.

Suppose you need to summarize a report.

Basic prompt:

Summarize this report.

Improved prompt:

ROLE:
Act as a careful document reviewer.

TASK:
Summarize the attached report
for someone with no technical background.

RULES:
- Use simple English.
- Write short sentences.
- Explain unfamiliar terms.
- Preserve important dates and numbers.
- Do not invent missing information.

OUTPUT:
Provide:
1. Main purpose
2. Three important findings
3. Key risks or limitations
4. Short conclusion

Include page references where possible.

We've combined role-based and constraint-based prompting.

The role guides the approach.

The constraints explain what matters.

The output section makes the result easier to review.

And because we ask for page references, we have a way to check the original document.

This is especially helpful when working with long reports, research papers, or contracts.

11. Your 5-Minute Exercise

Now it's your turn.

We'll practice making a complicated explanation easier to understand.

Step 1: Start with this paragraph

In order to facilitate the effective utilization of the online learning platform, students are required to complete the authentication process prior to accessing the educational resources.

It's a long way of saying something fairly simple.

Step 2: Ask AI to simplify it

Rewrite this paragraph in simple English.

Use short sentences.

Prefer familiar words.

Preserve the original meaning.

Do not add new information.

Paragraph:
[PASTE PARAGRAPH]

Step 3: Compare the result

A possible rewrite is:

Students must sign in before they can use the online learning materials.

It's shorter and easier to understand.

Notice that we didn't remove an important requirement: students still need to sign in first.

Step 4: Test one change

Try changing only one instruction.

For example:

Use numbered steps when the text describes a process.

Compare the new answer with the previous one.

Does the additional requirement make the explanation easier to follow?

Remember our testing habit from Part 2: save the original prompt and change one thing at a time.

Quick Recap

What's Coming in Part 5?

We've now learned how to write more structured prompts, choose useful patterns, and give clearer instructions.

But there's another question:

How do we know whether our prompt actually works?

In Part 5, we'll explore prompt evaluation.

We'll learn how to compare responses, identify common problems, and decide whether a prompt genuinely improves results.

We'll keep the examples simple and practical.

Final Thoughts

Role-based prompting tells AI how to approach a task.

Constraint-based prompting explains the rules it should follow.

When we combine both, we can guide the model toward more useful and consistent answers.

And ASD-STE100 teaches us something valuable:

Clear writing isn't just about using simpler words. It's about reducing confusion without losing meaning.

The same principle applies to prompt engineering.

You don't need complicated instructions to get started.

You need clear goals, relevant information, and rules you can check.

One Question Before You Go

Have you ever read an AI explanation that sounded impressive but was difficult to understand?

Try asking AI to rewrite it using short sentences and simple words.

Did the explanation become clearer?

Share your experience in the comments!

Part 4 of 6 in my "Prompt Engineering from Scratch" series. I work with and test open-source AI models on real documents, and I share practical lessons along the way.

Further Reading

• ASD-STE100 Official Website
• ASD-STE100 Official Standard and Downloads
• ASD-STE100 Frequently Asked Questions
• OpenAI Prompt Engineering Guide