Pair Programming with AI

Chat mode turns AI into a pair programming partner. You discuss, explore, debug, and design—through conversation.

Here’s how to make the most of it.

What Chat Mode Is For

Discussion and exploration: “How should I approach this problem?” “What are the trade-offs between these approaches?”

Debugging: “This is failing with error X. Here’s the stack trace…” “Why might this return null?”

Explanation: “Explain what this code does” “Why was this pattern chosen here?”

Code review: “Review this function for potential issues” “How could I make this more readable?”

Effective Prompts

Be specific about what you want:

Bad: “Help me with this code” Good: “Review this function for null handling issues”

Bad: “It’s not working” Good: “This returns undefined instead of the user object when called with id=5”

Provide Context

Include relevant code: “Here’s the function: [code]” “The error occurs at line 23”

Include error messages: “The full stack trace is: [trace]”

Include expected vs actual: “Expected: returns user object” “Actual: returns undefined”

Multi-Turn Conversations

Chat mode excels at back-and-forth:

Turn 1: “How should I implement caching here?” AI: Suggests approach Turn 2: “What about cache invalidation?” AI: Explains strategies Turn 3: “Can you show me how to implement the time-based approach?” AI: Provides code

Build on previous context.

Selecting Code

Highlight code before chatting:

1. Select relevant code in editor
2. Open chat
3. Ask question about selected code

AI knows exactly what you’re referring to.

Chat for Architecture

Discuss design decisions:

“I’m building a notification system. Users need to receive notifications via email, SMS, and in-app. How should I architect this?”

AI can help think through:

• Component structure
• Interface design
• Trade-offs
• Similar patterns in codebase

Chat for Debugging

Systematic debugging assistance:

Start: “This test is failing. Here’s the test and the function it’s testing.” AI: Analyzes, suggests potential issues You: “I checked that, it’s not the issue” AI: Suggests next hypothesis Continue until root cause found

AI remembers the conversation context.

Chat for Learning

Use chat to understand code:

“Walk me through what this function does step by step” “Why is this pattern used instead of [alternative]?” “What are the edge cases this handles?”

Learn while you work.

When to Use Chat vs. Copilot vs. Agent

Copilot: You’re typing, want suggestions Chat: You need to discuss, explore, or debug Agent: You want autonomous task execution

Chat when conversation helps. Switch when you need something else.

The Chat Mode Checklist

For effective chat-based coding:

• Clear question or request
• Relevant code included
• Error messages provided
• Expected vs actual stated
• Build on conversation history
• Know when to switch modes

Code through conversation.

Use chat mode in AI IDE →