Simple explanation
“Build the whole app” sounds like the fastest way to get started. In practice, it asks AI to make decisions about the product, screens, data, permissions, error states, and visual style in one pass.
When the result is wrong, it becomes difficult to know which assumption caused the problem. A large first prompt can also create a lot of code before you have tested whether the core user flow makes sense.
A better workflow is to build one small vertical slice: one real user action, from the first screen through the result. Then test it, learn from it, and add the next slice. You still move toward the full app, but every step stays understandable and recoverable.
What to do
Break the build into reviewable slices:
1. Write the one user problem the first version must solve.2. Choose the smallest complete flow, such as create an item, view it, and find it again.3. Ask AI to audit the existing project and propose the smallest next change before editing.4. Approve one feature at a time and state what must remain unchanged.5. Run the relevant checks and test the flow manually before adding another feature.6. Keep a known-good commit so you can recover if the next change goes wrong.
A slice is not a random fragment. It is a small path that a real user can complete and you can verify.
Copy-paste prompt
I am building [describe the app]. Do not build the whole app. First, help me choose the smallest complete user flow that tests the core value. Audit the project, list the files involved, explain the risks, and propose one safe change. Do not edit until I approve the plan. After editing, report the files changed, checks run, and what still needs manual testing.
Course note
Key takeaway
Do not ask AI to build the whole app in one go. Build one complete, testable user flow at a time.