How to Write an AI Prompt

Want to build your own app just by describing it? Vibe coding lets you generate functional software using natural language, though the precision of your prompt dictates the quality of your output. In this tutorial, we walk through building a fully playable 3D racing game using these core prompting strategies:

Be Granular and Specific: Open-ended prompts produce generic results. Break your request into structured bullet points or themes to define your vision.
Dictate Technical Keywords: Guide the model by choosing the underlying technology. Even if you aren’t technical, tell the model to find the best language or tools for your specific use case.
Be Granular and Specific: Open-ended prompts produce generic results. Break your request into structured bullet points or themes to define your vision.
Dictate Technical Keywords: Guide the model by choosing the underlying technology. Even if you aren’t technical, tell the model to find the best language or tools for your specific use case.
Include Visual References: Upload sketches, screenshots, or reference art alongside your text to establish the visual style and layout.

Iteration is a fundamental part of “vibe coding” because it allows you to refine your application until it aligns perfectly with your vision (1:55 – 2:04). Instead of settling for the first result produced by the AI, iteration offers several key advantages:

Performance Tuning: You can play the game you’ve built and identify specific areas that need improvement, such as how a car handles on a track or the duration of a boost (2:06 – 2:16).
Continuous Refinement: By going back to the prompt box to provide more granular instructions, you can steer the model toward the exact behavior or style you want (2:16 – 2:29).
Back-and-Forth Collaboration: Iteration creates a feedback loop between you and the model. This dialogue is essential for turning a generic result into a functional, polished piece of software that feels “just right” (2:29 – 2:35).

Writing a great AI prompt is less about “coding” and more about being a clear, effective manager. If you treat the AI like a highly capable but literal-minded intern, you’ll get much better results.

Here is a framework to take your prompts from generic to high-performance.

1. The Core Framework: R.T.C.F.

A professional-grade prompt usually contains these four elements:

Element	Purpose	Example
Role	Tell the AI who it should be.	“Act as a senior software architect…”
Task	Tell the AI exactly what to do.	“…write a technical documentation draft…”
Context	Give background and constraints.	“…for a new internal API aimed at junior devs.”
Format	Define the layout of the answer.	“Present this as a Markdown file with a table.”

2. Advanced Techniques

Once you have the basics down, use these “power user” tactics to refine the output:

Few-Shot Prompting (Give Examples):

Don’t just describe a style; show it.

“Write a product description in this style: [Example 1]. Now, write one for a [New Product].”
Chain-of-Thought (Think Step-by-Step):

For complex logic or math, ask the AI to explain its reasoning. This drastically reduces “hallucinations” (errors).

“Before giving the final answer, walk me through your logic step-by-step.”
Negative Constraints (The “Don’t” List):

Be explicit about what you don’t want to avoid clichés.

“Do not use corporate jargon or buzzwords like ‘synergy’ or ‘game-changer’.”
Iterative Refinement:

Treat the first response as a draft. You can follow up with: “Make the tone 20% more professional” or “Add a section specifically about the budget.”

3. Comparing Good vs. Bad Prompts

❌ The “Lame” Prompt:

“Write a blog post about AI.”

(Result: A generic, boring 500-word essay you’ve seen a thousand times.)

✅ The “Pro” Prompt:

“Act as a tech journalist for Wired. Write a 400-word blog post about the rise of agentic AI in 2026. Focus on how it impacts small business owners. Use an engaging, slightly skeptical tone. End with three actionable tips for non-technical founders.”