← Back to Learn
Best PracticesIntermediate10 min read

Humanizing Your LLM Prompts

Techniques to make AI assistants write code that feels more natural and less robotic.

Garbage in, garbage out. If you ask an LLM to "write a function," it will give you the most generic, textbook version of that function. To get "soulful" code, you need to ask for it.

The "Vibe" Prompt

Don't just describe *what* you want. Describe *how* you want it.

**Bad Prompt:**

"Write a React component for a user card."

**Good Prompt:**

"Write a React component for a user card. Use Tailwind CSS. Keep it minimal and pragmatic. Avoid unnecessary comments. Use destructuring. Assume the parent component handles loading states."

Key Keywords to Use

**"Pragmatic"**: Tells the AI to avoid over-engineering.
**"Idiomatic"**: Encourages language-specific best practices (e.g., Pythonic code).
**"Minimal"**: Reduces boilerplate and verbose comments.
**"Senior Developer"**: Sets a persona that implies cleaner, more maintainable code.

Example Workflow

1. **Generate**: Ask for the initial code.
2. **Critique**: "That looks too robotic. Remove the JSDoc comments and use arrow functions."
3. **Refine**: "Better. Now rename `data` to `userProfile` to match our domain."

By treating the LLM as a junior developer who needs mentorship, you can guide it toward producing code that feels like it was written by a human.

Up Next

Comment Quality Over Quantity

Read Guide →
Vibe Code Detector