M3-A - What is a Prompt?

Start With the Wrong Mental Model

Beginners usually treat prompting like talking:

• "please explain this"
• "make this better"
• "write something good"

That works for casual use, but it breaks once the task matters.

Because the model is not a mind-reader.

It does not know:

• what level you want
• who the answer is for
• what "better" means
• what shape the output should take

So the first upgrade is this:

A prompt is not just a message to AI. A prompt is the environment that defines the job.

What a Prompt Actually Does

A good prompt quietly answers five questions:

So when you write:

Explain operating systems to a first-year BSc CS student in simple bullets and include one real-world analogy.

you are not only requesting an answer.
You are defining a target.

That target gives the model less room to wander.

Prompting Is About Reducing Guesswork

The model is always trying to predict what kind of answer should come next.

If your prompt is vague, the model must guess more.

If your prompt is sharp, the model has a narrower problem to solve.

That is why these two prompts feel so different:

Prompt 1

Tell me about AI.

Prompt 2

Explain AI to a 16-year-old student who knows basic computers but not coding. Use one school-life example, 6 bullet points, and end with a 3-line recap. Avoid jargon.

Prompt 2 is not "more polite."
It is more precise.

Prompt = Task + Context + Constraints

You can remember prompting through a simple engineering lens:

Task

What exactly should the model do?

• explain
• rewrite
• compare
• summarize
• critique
• generate questions

Context

What situation should the model know?

• who the audience is
• what they already know
• where the text came from
• what the goal is

Constraints

What boundaries matter?

• use bullets
• keep under 120 words
• no jargon
• do not invent facts
• use only the notes provided

This is why strong prompting feels less like chatting and more like lightweight system design.

Why Prompting Feels Easy at First and Hard Later

At beginner level, prompting feels simple because:

• the model is forgiving
• basic prompts often still produce usable output

At advanced level, prompting becomes harder because:

• consistency matters
• format matters
• edge cases matter
• hallucination matters
• reuse matters

So the deeper skill is not "how to write fancy prompts."

The deeper skill is:

how to define behavior clearly enough that output becomes reliable.

Prompting vs Coding

Traditional code gives explicit instructions.

Prompting gives structured language instructions to a probabilistic system.

This is why prompts are powerful but slippery.

You are not writing strict logic.
You are shaping probability.

A More Useful Definition

Instead of saying:

• "A prompt is what we type."

say:

• "A prompt is the specification that tells the model what kind of answer space to operate inside."

That definition sounds heavier, but it is much more accurate.

And once you see it that way, you naturally start writing better prompts.

Three Levels of Prompting

Level 1: Casual prompting

Example:

Explain photosynthesis.

Useful for fast help. Weak for precision.

Level 2: Directed prompting

Example:

Explain photosynthesis to a 10th standard student using one plant analogy and 5 bullet points.

Much stronger for practical use.

Level 3: Designed prompting

Example:

You are a biology tutor. Explain photosynthesis to a 10th standard student who struggles with science vocabulary. Use one everyday analogy, 5 bullet points, one "common confusion" section, and a 2-line recap. Avoid jargon and keep it under 180 words.

This is where prompting starts becoming a real skill.

Real Examples

Example 1: Notes

Weak:

Make notes on networking.

Stronger:

Create revision notes on computer networking for a first-year BSc CS student. Use short headings, a small comparison table, and 5 viva questions at the end.

Example 2: Writing

Weak:

Improve this paragraph.

Stronger:

Rewrite this paragraph to sound clearer and more confident for a college seminar. Keep the meaning same, shorten repetition, and keep it under 100 words.

Example 3: Studying

Weak:

Help me study AI.

Stronger:

Turn these AI notes into a one-page revision sheet for a 12th standard student. Keep only high-yield ideas, use bullets, and add 5 self-test questions.

The Practical Rule

If the output matters, your prompt should usually define:

1. the task
2. the audience
3. the format
4. the constraints
5. the context

Miss two or three of these, and the model starts improvising.

Sometimes improvisation is useful.
Sometimes it is exactly what you do not want.

Final Frame

The beginner asks:

• "What should I type?"

The stronger user asks:

• "What conditions would make the right answer easier for the model to produce?"

That second question is the start of serious prompting.

Recap