Your First Session

You’ve picked a tool and set up your environment. Let’s write some code.

Step 1: Open Your Tool’s AI Surface

• in an IDE or extension, open the chat, command palette, or edit panel
• in a CLI tool, start the tool in your terminal and type your prompt directly

If you need product-specific shortcuts, use the vendor pages in the Reference Appendix.

Step 2: Start with Exploration

Do not start by commanding. Start by exploring. Your first prompt should be something you can verify:

Look at this codebase and explain:
1. What is the main entry point?
2. How is the code organized?
3. What patterns and conventions are used?

Why this works: you are asking about something you already know, or can check quickly. That is how you learn when the AI is helping and when it is bluffing.

Step 3: Ask for a Small Change

Pick something simple that you know how to do manually:

Add a utility function to src/utils.ts that validates email addresses.
Use the same style as the other functions in that file.

Good first tasks:

• Add a utility function
• Write a test for an existing function
• Add TypeScript types to an untyped function
• Refactor a function to be more readable

Bad first tasks:

• “Build me a login system”
• “Refactor the entire codebase”
• Anything you couldn’t do yourself

Step 4: Review the Output

Before accepting any changes:

1. Read the diff: understand what changed
2. Check for errors: does the LSP show problems?
3. Run tests: do existing tests still pass?
4. Verify behavior: does it do what you asked?

Never blindly accept

If you don’t understand the code, don’t ship it. Ask the AI to explain, or reject the change.

Step 5: Iterate

If something’s wrong, be specific:

The email validation doesn't handle plus signs in the local part.
Update it to allow addresses like user+tag@example.com.

Good correction:

“The function doesn’t handle X. Fix it by doing Y.”

Bad correction:

“Fix it” (too vague) “That’s wrong” (not actionable)

The Golden Rule in Practice

You asked for email validation because you know what correct email validation looks like. This lets you:

• Spot if the regex is wrong
• Notice if edge cases are missing
• Verify the code actually works

Once you can reliably spot mistakes on familiar tasks, you’re ready for unfamiliar territory.

Common First-Session Mistakes

Mistake	Fix
Too big a first ask	Start smaller, one function, not a feature
Accepting without reading	Always review diffs before accepting
No verification	Run linter/tests after every change
Giving up too fast	Try 2-3 iterations before abandoning

What’s Next

You’ve completed your first AI-assisted coding session. To work more effectively:

• Learn the core concepts: terminology and mental models
• Study effective patterns: what actually works
• Avoid common mistakes: save yourself pain