Apply it · Prompting

Build a prompt you can use.

Start with the task you care about. Add the context, boundaries, and format that help AI produce useful work. Your prompt is assembled as you go.

Step 1 of 8

Step 1 · Objective

What do you want AI to accomplish?

A strong prompt starts with the result you want, not with a long role description. Be specific about the task and the desired outcome.

Why this matters

The AI needs a clear destination. Vague goals force it to guess what success looks like.

Describe the outcomeRecommended

Start with a verb such as create, analyze, compare, explain, debug, rewrite, plan, or evaluate.

BetterSpecific

Create a production-ready SaaS landing page that explains the product clearly and increases demo requests.

WeakerVague

Make me a good website.

Step 2 · Perspective

Does the task benefit from a role?

Roles can guide perspective and standards, but they should be used only when they materially improve the work.

Why this matters

A useful role adds domain judgment. An inflated role usually adds noise rather than quality.

Optional role or perspectiveOptional

Leave this blank when the task does not need a specialist perspective.

BetterSpecific

Act as a senior product designer and frontend developer.

WeakerVague

Act as the world's greatest genius AI guru with 100 years of experience.

Step 3 · Context

What does the AI need to know before starting?

Context is the information that changes how the task should be solved: project state, audience, business objective, available data, or environment.

Why this matters

Without relevant context, the AI may produce a technically correct answer that is wrong for your actual situation.

Relevant backgroundRecommended

Include only details that can change the answer. More context is not automatically better.

BetterSpecific

The app already exists in Next.js. Keep the current APIs and brand. Target users are small SaaS teams.

WeakerVague

Here is every detail about the company since 2014...

Step 4 · Requirements

What must the result include?

Requirements define the non-negotiable features, qualities, facts, or behaviors that make the result usable.

Why this matters

Requirements turn your idea of good into concrete acceptance criteria the AI can follow.

Must-have requirementsRecommended

Focus on things you would check before accepting the result.

BetterSpecific

Responsive layout; accessible forms; clear pricing section; production-quality code.

WeakerVague

Make it awesome, modern, beautiful, amazing, and perfect.

Step 5 · Constraints

What should the AI not change or violate?

Constraints protect the parts of the problem that must stay fixed: technologies, limits, policies, formats, budgets, architecture, or exclusions.

Why this matters

Good boundaries reduce rework. They prevent the AI from solving the task by changing assumptions you needed to preserve.

Boundaries and exclusionsRecommended

Prefer concrete restrictions over broad instructions such as do not make mistakes.

BetterSpecific

Do not change existing APIs. Do not add paid dependencies. Keep the current brand colors.

WeakerVague

Do not do anything bad or wrong.

Step 6 · Method

How should the AI approach the task?

For complex work, tell the AI what useful actions to perform before delivering the result: inspect, compare, test, verify, calculate, or validate.

Why this matters

Process instructions are valuable when the task requires disciplined work. Avoid ceremonial steps that do not improve the result.

Useful working approachRecommended

Ask for observable work such as testing, checking, comparing, or verifying.

BetterSpecific

Inspect the implementation, identify gaps, implement the solution, then test affected flows.

WeakerVague

Think for exactly 27 steps and never stop thinking.

Step 7 · Deliverable

What should the final answer look like?

Specify the deliverable, structure, level of detail, and format when they matter. This makes the response easier to use immediately.

Why this matters

An AI can solve the right problem but still return it in the wrong shape. Output instructions reduce that mismatch.

Expected outputRecommended

Mention format only when it helps: code, table, checklist, JSON, report, concise answer, etc.

BetterSpecific

Return the completed implementation, then a short change summary and important assumptions.

WeakerVague

Give me the answer in the best possible format.

Step 8 · Review

Review the finished prompt before you use it.

The best prompt is not the longest prompt. It is the minimum sufficient instruction needed for the AI to understand the task and produce the right result.

Why this matters

Before copying, check for contradictions, duplicated instructions, vague adjectives, missing context, and requirements that cannot be verified.

Optional final noteOptional

Use this only for an important final rule that does not fit elsewhere.

BetterSpecific

If a necessary assumption cannot be verified, state it explicitly.

WeakerVague

Make absolutely sure this is the best answer ever made by any AI.

Apply it · Prompting

Build a prompt you can use.

Start with the task you care about. Add the context, boundaries, and format that help AI produce useful work. Your prompt is assembled as you go.

Step 1 of 8

Step 1 · Objective

What do you want AI to accomplish?

A strong prompt starts with the result you want, not with a long role description. Be specific about the task and the desired outcome.

Why this matters

The AI needs a clear destination. Vague goals force it to guess what success looks like.

Describe the outcomeRecommended

Start with a verb such as create, analyze, compare, explain, debug, rewrite, plan, or evaluate.

BetterSpecific

Create a production-ready SaaS landing page that explains the product clearly and increases demo requests.

WeakerVague

Make me a good website.

Step 2 · Perspective

Does the task benefit from a role?

Roles can guide perspective and standards, but they should be used only when they materially improve the work.

Why this matters

A useful role adds domain judgment. An inflated role usually adds noise rather than quality.

Optional role or perspectiveOptional

Leave this blank when the task does not need a specialist perspective.

BetterSpecific

Act as a senior product designer and frontend developer.

WeakerVague

Act as the world's greatest genius AI guru with 100 years of experience.

Step 3 · Context

What does the AI need to know before starting?

Context is the information that changes how the task should be solved: project state, audience, business objective, available data, or environment.

Why this matters

Without relevant context, the AI may produce a technically correct answer that is wrong for your actual situation.

Relevant backgroundRecommended

Include only details that can change the answer. More context is not automatically better.

BetterSpecific

The app already exists in Next.js. Keep the current APIs and brand. Target users are small SaaS teams.

WeakerVague

Here is every detail about the company since 2014...

Step 4 · Requirements

What must the result include?

Requirements define the non-negotiable features, qualities, facts, or behaviors that make the result usable.

Why this matters

Requirements turn your idea of good into concrete acceptance criteria the AI can follow.

Must-have requirementsRecommended

Focus on things you would check before accepting the result.

BetterSpecific

Responsive layout; accessible forms; clear pricing section; production-quality code.

WeakerVague

Make it awesome, modern, beautiful, amazing, and perfect.

Step 5 · Constraints

What should the AI not change or violate?

Constraints protect the parts of the problem that must stay fixed: technologies, limits, policies, formats, budgets, architecture, or exclusions.

Why this matters

Good boundaries reduce rework. They prevent the AI from solving the task by changing assumptions you needed to preserve.

Boundaries and exclusionsRecommended

Prefer concrete restrictions over broad instructions such as do not make mistakes.

BetterSpecific

Do not change existing APIs. Do not add paid dependencies. Keep the current brand colors.

WeakerVague

Do not do anything bad or wrong.

Step 6 · Method

How should the AI approach the task?

For complex work, tell the AI what useful actions to perform before delivering the result: inspect, compare, test, verify, calculate, or validate.

Why this matters

Process instructions are valuable when the task requires disciplined work. Avoid ceremonial steps that do not improve the result.

Useful working approachRecommended

Ask for observable work such as testing, checking, comparing, or verifying.

BetterSpecific

Inspect the implementation, identify gaps, implement the solution, then test affected flows.

WeakerVague

Think for exactly 27 steps and never stop thinking.

Step 7 · Deliverable

What should the final answer look like?

Specify the deliverable, structure, level of detail, and format when they matter. This makes the response easier to use immediately.

Why this matters

An AI can solve the right problem but still return it in the wrong shape. Output instructions reduce that mismatch.

Expected outputRecommended

Mention format only when it helps: code, table, checklist, JSON, report, concise answer, etc.

BetterSpecific

Return the completed implementation, then a short change summary and important assumptions.

WeakerVague

Give me the answer in the best possible format.

Step 8 · Review

Review the finished prompt before you use it.

The best prompt is not the longest prompt. It is the minimum sufficient instruction needed for the AI to understand the task and produce the right result.

Why this matters

Before copying, check for contradictions, duplicated instructions, vague adjectives, missing context, and requirements that cannot be verified.

Optional final noteOptional

Use this only for an important final rule that does not fit elsewhere.

BetterSpecific

If a necessary assumption cannot be verified, state it explicitly.

WeakerVague

Make absolutely sure this is the best answer ever made by any AI.