Use AI to write code you actually understand
A beginner’s guide to asking better coding questions, reviewing changes, testing results, and protecting your secrets.
Give the assistant a small, testable task
Start with a result you can check: add a button that changes the heading, or explain why a form reloads the page. Describe the current behavior, the intended behavior, and the files involved. Asking for a complete business app before you understand its parts makes mistakes harder to find.
A prompt you can adapt
Tell the assistant your level, show the smallest relevant example, and describe what you tried. Ask it to explain the proposed change before applying it. Do not include passwords, private customer data, API keys, or payment credentials.
I am learning JavaScript.
My button should increase a counter by one.
Instead, the whole page reloads.
Here is the relevant HTML and JavaScript: [paste your example]
Explain the cause and suggest the smallest change.
Then give me three checks to confirm it works.Read, run, test, explain
Read the changed code. Run it locally. Test the ordinary case, an empty input, and a repeated action. Then explain the change without looking at the assistant’s explanation. If you cannot, ask a follow-up question or make the example smaller. Generated code is a proposal that still needs review.
Keep a working version
Commit a working version with Git before asking for a larger change. Review the difference afterward and undo changes you do not understand. Avoid changing the UI, database, and authentication in one prompt. Small changes let you identify which decision introduced a problem.
Protect access and control cost
Keep secret keys on the server and outside Git. Do not treat hidden buttons as permission checks: a server must decide who may read or change private data. Set usage limits for paid APIs, check the current provider terms, and never assume a free tier is unlimited.
Practice both with and without assistance
Build a small feature by hand, then ask AI for an alternative. Compare readability and behavior. The goal is to make better decisions with help while retaining the ability to understand and fix the result. Your project portfolio should show that understanding, including what failed and how you corrected it.