AI as Pair Programmer

Time: ~5 minutes | Difficulty: Beginner

What You'll Learn

• The right mental model for AI collaboration
• What AI is good at (and not)
• How to think about AI as a tool

The Big Idea

AI is not a magic button that writes code for you. It's more like having a very knowledgeable colleague sitting next to you — one who knows a lot but still needs clear instructions and sometimes makes mistakes.

The Right Mental Model

❌ Wrong: AI as Oracle

"Just tell AI what you want and it'll figure it out perfectly."

✅ Right: AI as Pair Programmer

"Work together, where I drive and AI assists."

In pair programming:

• One person is the Driver (types, makes decisions)
• One person is the Navigator (suggests, reviews, catches errors)

With AI:

• You are always the Driver
• AI is the Navigator

What AI Is Good At

Strength	Example
Generating boilerplate	"Create a React component with these props"
Explaining concepts	"Explain what useEffect does"
Debugging	"Why might this error be happening?"
Suggesting patterns	"What's a good way to structure this?"
Translation	"Convert this to TypeScript"
Documentation	"Add comments explaining this code"

What AI Is Not Good At

Limitation	Reality
Reading minds	It can't know your full context without you explaining
Perfect code	It makes mistakes, uses outdated patterns
Business logic	Only you know what your app should do
Security guarantees	Always verify security-sensitive code
Running code	It can write code, but you must run and test it

The Amplification Effect

Here's the key insight:

AI amplifies your current abilities.

• If you vaguely understand React, AI helps you learn faster
• If you clearly understand what you want, AI builds it faster
• If you're confused, AI can't read your mind

The more you understand, the better your prompts, the better the output.

A Day in Pair Programming

Without AI:

1. Think about what to build
2. Look up documentation
3. Write code slowly
4. Debug errors
5. Repeat

With AI:

1. Think about what to build      ← Still you
2. Ask AI for approach            ← AI helps
3. AI writes initial code         ← AI generates
4. You review and understand      ← Still you
5. You run and test               ← Still you
6. Fix issues together            ← Both

Notice: You're still thinking, reviewing, and running. AI accelerates the middle parts.

The Trust Spectrum

How much should you trust AI code?

Blind Trust     Reasonable Trust     Verify Everything
❌ ─────────────────── ✓ ────────────────────── ❌
    Never              Sweet spot           Too slow

Reasonable trust means:

• Read the code before running
• Understand what it does
• Test that it works
• Be extra careful with security

Real Example

Bad approach:

"Build my app"
→ AI produces something, you have no idea what it does
→ You can't debug it because you don't understand it

Good approach:

"Create a login form component with email and password fields"
→ AI produces focused code
→ You understand what you asked for
→ You can verify it does what you wanted

Your Responsibilities

When working with AI, you're responsible for:

1. Clear instructions — AI can't read your mind
2. Understanding — Don't commit code you don't understand
3. Verification — Run the code, test it works
4. Security — Never share passwords, API keys
5. Final decisions — You're the author, not AI

Check Your Understanding

• [ ] AI is a pair programmer, not a magic button
• [ ] I'm the Driver; AI is the Navigator
• [ ] AI amplifies my abilities
• [ ] I need to verify and understand AI code
• [ ] Clear instructions get better results

Next Up

Now let's learn how to write prompts that get good results.

Continue: Anatomy of a Good Prompt →