Blog / Build with AI /

How to learn building with AI in thirty days

Most people who want to learn building with AI do it backwards. They watch videos for a month, collect tools, and build nothing.

Thirty days is enough to go from curious to someone who has shipped small things, if you spend them making. Here is a plan built around finishing one project each week. It assumes about 45 to 60 minutes a day. If you have less, stretch it to sixty days. The order matters more than the speed.

You do not need to be a developer. You do need to be willing to read code, break things and try again.

Before day one: set up

You need only three things:

  • An AI assistant. Any of the main chat assistants will do. Use one so you learn its habits.
  • A text editor. Visual Studio Code is free and common, but Notepad works for week one.
  • A folder for your projects, and a note file called log.md. Every day, write two lines: what you tried, what surprised you.

The log is the secret ingredient. It shows progress when it feels like there is none, and it becomes your own reference.

Week 1 (days 1 to 7): one file, one job

Goal: build and open a single-page tool in your browser.

  • Day 1. Read choosing your first AI project and pick yours. Write the one-sentence definition of done.
  • Day 2. Ask the AI for the smallest working version. Open it in a browser. Do not edit anything yet.
  • Day 3. Ask it to explain the code, section by section. Highlight the three parts you understand and the three you do not.
  • Day 4. Change something small yourself: a colour, a label, a number. Break it on purpose, then fix it.
  • Day 5. Test it with awkward input: empty, huge, strange characters.
  • Day 6. Fix what you found, one change per request. Small changes beat big rewrites.
  • Day 7. Stop. Use it once for real. Write what you would change.

Week 2 (days 8 to 14): learn the vocabulary

Goal: be able to describe what you want so the first draft is closer.

  • Day 8. Read how to describe a tool to AI so it builds the right thing. Rewrite your prompt from week one using it.
  • Day 9. Learn four words: HTML (structure), CSS (looks), JavaScript (behaviour), and a function. Ask the AI to show each in your project.
  • Day 10. Use your browser’s developer tools. Right-click, Inspect, and change text on a live page to see what happens.
  • Day 11. Read my checklist before asking AI to build anything and make your own version.
  • Day 12. Start project two, a bit more ambitious: a tool with two inputs, or one that remembers a setting in the browser.
  • Day 13. Build, test, fix.
  • Day 14. Finish it. Compare your prompts with day two’s. You should see the difference.

Week 3 (days 15 to 21): safety and honesty

Goal: learn where AI-built things go wrong.

  • Day 15. Read AI hallucinations in code and then hunt for an invented function in your own project. Ask the AI directly whether anything it used is not real.
  • Day 16. Learn Git basics from Git basics for non-developers. Put your two projects under version control.
  • Day 17. Read security basics for AI-built web apps. Check whether either project handles user input.
  • Day 18. Build a small browser extension or bookmarklet. Building a browser bookmarklet with AI in ten minutes is a good route.
  • Day 19. Add one test for each project: a list of five inputs and the results you expect.
  • Day 20. Put one project online with free static hosting.
  • Day 21. Share the link with one friend. Watch them use it without helping. Take notes.

Week 4 (days 22 to 30): build something you will keep

Goal: a useful tool that solves a real problem of yours.

  • Day 22. Pick the problem. Aim for a small repeated task from your work. If you manage websites, look for a chore you do weekly.
  • Day 23. Write a short brief: users, inputs, outputs, what is out of scope.
  • Day 24. Get a first version running. Keep the stack boring.
  • Day 25. Test it with real data from your work.
  • Day 26. Fix three things. Not ten.
  • Day 27. Add a simple help line or instruction on the page.
  • Day 28. Use it for a full day of real work.
  • Day 29. Write a one-page note on how it works. Include what you do not understand yet.
  • Day 30. Review your log. List what you can do now, what still confuses you and what to build next.

Rules that make the plan work

  1. Finish before you start something new. An unfinished project teaches less than a small finished one.
  2. Read what the AI gives you, even if only to find the confusing parts.
  3. Change one thing at a time.
  4. Never paste passwords, API keys or private customer data into a chat.
  5. Keep tools boring for the first month. Frameworks can wait.
  6. Ask “why?” as often as “how?”.

What to ignore

  • Endless tool comparisons. Pick one assistant for the month.
  • Huge courses. They can help later; they are a poor first step.
  • Other people’s big launches. Their path is not your plan.
  • The feeling that you should already know this.

When it gets hard

Around day 12 and again around day 20 most learners hit a wall. A bug the AI cannot fix. A concept that will not click. The answer is almost always the same: shrink the problem. Remove features until it works, then add back one at a time. If you are stuck for more than an hour, write down exactly what you expected and what happened, and start a fresh chat with only that.

Key takeaways

  • Learn by finishing: one small project each week.
  • Week one is one file; week two is better prompts; week three is safety; week four is something you will keep.
  • Keep a short daily log.
  • Shrink the problem when you are stuck.

Which day of this plan would you struggle with most, and what could you do this week to make it easier?