Skip to main content

​
How iteration works

Send a follow-up message describing what you want to change. The AI makes targeted code updates — it doesn’t rebuild from scratch. Each change follows this cycle:
  1. You describe the change
  2. AI modifies the relevant code
  3. App rebuilds automatically
  4. Preview updates with your changes
  5. A version checkpoint is saved

​
Types of changes

Design changes
  • “Change the background color to dark gray”
  • “Make the cards have rounded corners and a subtle shadow”
  • “Use a larger font for the page title”
Feature additions
  • “Add a settings screen with a dark mode toggle”
  • “Add pull-to-refresh to the home screen”
  • “Add a search bar that filters the list”
Screen additions
  • “Add a profile page that shows the user’s name and email”
  • “Add an onboarding flow with 3 intro slides”
Bug fixes
  • “The save button doesn’t do anything when tapped”
  • “Items overlap on smaller screens”
  • “The back button goes to the wrong screen”
Configuration
  • “Change the app name to FitTracker”
  • “Update the app icon”

​
Tips for effective iteration

One change at a time

Focused messages produce better results than bundling multiple changes together.

Be specific

“Make the header 60px tall with white text” is better than “fix the header.”

Reference screens

“On the home screen, move the search bar above the list” tells the AI exactly where to look.

Describe the problem

For bugs, describe what happens vs. what you expected. The AI will diagnose and fix it.

​
Version history

Every change creates a version checkpoint. You can:
  • Browse history — see what changed at each step
  • Roll back — revert to a previous version if a change didn’t work out
If a change goes wrong, you can always roll back. Don’t hesitate to experiment.