I Had AI Teach Me to Code for 30 Days (Without a Computer Science Degree)

I am not a programmer.

A month ago, I could barely write an Excel formula without Googling it. The closest I ever got to coding was changing the color of my MySpace profile in 2007 by copy-pasting HTML I did not understand.

But I had a problem that needed solving.

I was paying a developer $1,400 to build a simple inventory tracker for my side business. It was taking forever. Every change request added another $200 and another week to the timeline. I was hemorrhaging money for what felt like a spreadsheet with buttons.

So I asked myself a stupid question: Could AI just teach me to build it myself?

Thirty days later, I had built three working apps. The inventory tracker that would have cost me $1,400. A budget calculator I was going to pay $600 for. And a little tool that auto-generates social media captions that I was about to spend $2,200 on from Fiverr.

Total saved: $4,200.

Total time invested: About 60 hours over 30 days, mostly at night after work.

Here is exactly what happened.

Day 1-3: I Had No Idea What I Was Doing

The first thing I did was ask ChatGPT a embarrassingly basic question:

"I want to build a simple app that tracks inventory for my business. I have never coded before. Where do I start?"

ChatGPT told me I needed to learn HTML, CSS, and JavaScript. It gave me a roadmap. It said I should start with HTML basics and build up from there.

That sounded like a lot.

So I asked a better question:

"Can you teach me to build this app step by step, explaining everything like I'm 12 years old? Start with the absolute first thing I need to do."

That worked.

It told me to open a text editor (I used Notepad on Windows), create a file called index.html, and paste in 10 lines of code. It explained what each line did in plain English.

I did it. I opened the file in Chrome. It worked. It said "My First Web Page" in giant letters.

I felt like a genius.

For the next two days, I just followed instructions. ChatGPT would give me a block of code. I would paste it. It would break. I would copy the error message back to ChatGPT. It would fix it. I would paste the new code. It would work.

I had no idea what I was doing, but I was making progress.

By day 3, I had a webpage with a button that said "Add Item" and a list that showed fake inventory items. It did not save anything. It did not connect to a database. But it looked like a real app.

That was the moment I realized this might actually work.

Day 4-7: I Started Understanding What the Code Actually Did

The first three days were just copy-paste. I did not understand anything. I was a human Ctrl+V machine.

But around day 4, something clicked.

I started recognizing patterns. I saw the same structures repeating. When ChatGPT said "this is a function," I started to get what that meant. When it said "this is a loop," I could see how it repeated the same action multiple times.

I stopped just copying. I started tweaking.

ChatGPT would give me code to add a button. I would change the button text. Then I would change the color. Then I would add a second button and see if I could make it do something different.

Half the time, I broke everything. But I was learning faster by breaking stuff and fixing it than I ever did by just following instructions.

By day 7, I had a working inventory tracker that let me add items, delete items, and see a running total of quantities. It still did not save anything (if I refreshed the page, everything disappeared), but it felt real.

I showed it to a friend who codes professionally. He said it looked like something a bootcamp student would build in week 2.

I took that as a huge compliment.

Day 8-12: The Part Where I Almost Quit

This was the hard part.

Up until now, everything was visual. I could see buttons and lists and colors. It was satisfying.

But to make the app actually useful, I needed it to save data. That meant learning about something called "local storage" and eventually connecting to a real database.

I asked ChatGPT how to do it.

It gave me instructions. I followed them. Nothing worked.

I spent two entire evenings debugging one error. The app would not save data. ChatGPT kept giving me fixes that did not fix anything. I was pasting error messages and getting responses that made no sense.

I genuinely thought about quitting.

Then I switched to Claude.

I do not know why I did not think of this earlier, but I copy-pasted my entire broken code into Claude and said:

"This is supposed to save inventory data to local storage, but it's not working. I've been stuck for two days. Can you debug this and explain what's wrong in simple terms?"

Claude found the problem in 30 seconds. I had a typo. One letter. The variable was called itemName in one place and ItemName in another. Capitalization mattered. I had no idea.

Claude fixed it, explained why it broke, and gave me a tip: always use consistent naming.

The app worked.

I almost cried.

Lesson learned: If one AI gets stuck, try a different one. They explain things differently. Sometimes that is all you need.

Day 13-18: I Built My Second App Way Faster

After the inventory tracker was working, I felt unstoppable.

I had a budget calculator I had been planning to hire someone to build. It was supposed to take my income, subtract expenses, and show me how much I could save each month.

Instead of hiring someone, I just asked Claude:

"I want to build a budget calculator web app. I already know basic HTML, CSS, and JavaScript from building an inventory tracker. Walk me through building this."

Claude gave me a plan. I built it in 4 days.

Four. Days.

The first app took me 12 days to get to a working state. The second one took 4.

Why? Because I was not learning syntax anymore. I was learning logic.

I knew how to make a button. I knew how to make a form. I knew how to do math on user inputs. I just needed Claude to help me structure it correctly.

The budget calculator worked. It saved data. It had a clean interface. I showed it to my partner and she actually started using it.

That felt better than saving the $600 I would have paid a developer.

Day 19-25: I Built Something I Didn't Even Plan to Build

By week 3, I was dangerous.

Not good. But dangerous.

I was confident enough to try building things just to see if I could.

I run a small Instagram account for my side business. Writing captions takes forever. I had gotten a quote from a Fiverr freelancer to build a tool that would generate captions based on a photo description. They wanted $2,200.

I thought: I bet I could build that myself now.

I asked ChatGPT:

"I want to build a web app that takes a text description of a photo and generates 5 Instagram caption options. I want to use the OpenAI API to generate the captions. Walk me through it step by step."

ChatGPT walked me through it.

I had to sign up for an OpenAI API key (costs money per use, but like $0.02 per caption). I had to learn how to make an API call from my app. I had to handle errors when the API was slow or failed.

It took me 6 days. But it worked.

I type in "photo of a latte with foam art," click generate, and I get 5 caption options in 10 seconds.

I use it multiple times a week now. It has already saved me hours of writing time.

Cost to build: $0 (plus about $3 in API usage over the last week).

Cost if I had hired it out: $2,200.

Day 26-30: I Started Teaching Someone Else

By the end of the month, my friend who saw my week 1 inventory tracker asked me how I learned so fast.

I told him: I did not learn to code. I learned how to use AI as a coding tutor.

He wanted to try it.

I sat with him for 30 minutes and walked him through the same process. Ask AI for a roadmap. Build something simple first. Break it. Fix it. Ask better questions when you get stuck. Switch AIs if one is not helping.

Within an hour, he had a working webpage with a button that changed the background color when you clicked it.

He was hooked.

Watching him go through the same "holy shit, this actually works" moment I had on day 3 made me realize how repeatable this process is.

You do not need to be technical. You do not need a computer science degree. You do not even need to understand what the code does at first.

You just need to be willing to copy, paste, break, and fix things for a few weeks until patterns start to make sense.

What I Actually Learned (Besides Code)

Coding was the output. But the real skill I learned was how to collaborate with AI.

Most people treat AI like Google. They ask a question, get an answer, and move on.

That does not work for learning.

What worked for me was treating AI like a patient tutor who I could ask the same question 15 different ways until I understood it.

When ChatGPT gave me code I did not understand, I would paste it back and say:

"Explain this line by line like I'm in 5th grade."

When Claude fixed a bug, I would ask:

"Why did that fix it? What was wrong with my original code?"

When I built something that worked but looked ugly, I would ask:

"How do I make this look more professional without learning CSS frameworks?"

The AIs never got frustrated. They never made me feel dumb. They just kept explaining until I got it.

That is the unlock.

If you are willing to ask follow-up questions and admit when you do not understand, AI will teach you almost anything.

The Tools I Actually Used

Here is the full toolkit I used over 30 days:

AI Tools:

• ChatGPT (free version): My main teacher. Best for step-by-step instructions and explaining concepts.
• Claude (free version): Better at debugging and finding errors in my code. I used this when ChatGPT got stuck.
• Perplexity (free): I used this a few times to research specific concepts when I wanted a second explanation.

Code Tools:

• Notepad (Windows built-in): Where I wrote my code for the first week. Zero learning curve.
• VS Code (free): I switched to this in week 2 because ChatGPT recommended it. It color-codes your code and catches some errors automatically. Made life way easier.
• Chrome Browser: For testing my apps. Right-click → Inspect lets you see errors in real-time.

Total cost: $0.

Everything I used was free.

What Worked Really Well

1. Starting with a real project instead of tutorials

I did not do a single coding tutorial. I did not watch a single YouTube video.

I just told AI what I wanted to build and started building it.

That kept me motivated. Every day I could see progress. If I had spent two weeks learning syntax before building anything, I would have quit.

2. Breaking things on purpose

Once I got something working, I would intentionally change parts of the code to see what would break.

That taught me what each piece actually did.

3. Using multiple AIs

ChatGPT, Claude, and Perplexity all explain things slightly differently.

When I got stuck on a concept, I would ask all three. One of them would always explain it in a way that clicked.

4. Building the same thing twice

After I finished the inventory tracker, I rebuilt it from scratch without looking at the old code.

The second version took 90 minutes instead of 12 days. And I understood every line.

What Did Not Work

1. Asking vague questions

Early on, I would ask things like "How do I make this better?"

That is useless. AI does not know what "better" means to you.

I learned to ask specific questions: "How do I make this button bigger?" "How do I center this text?" "How do I make the list scroll if there are more than 10 items?"

Specific questions get useful answers.

2. Copying code I did not understand and moving on

When I just pasted code and moved to the next step without understanding what it did, I would hit a wall later.

I forced myself to ask "What does this line do?" for anything that looked confusing.

That slowed me down at first, but I learned way faster.

3. Trying to learn everything before building

I wasted the first two days reading articles about "what programming language should I learn first?"

Waste of time.

Just pick a project and start. You will learn what you need as you go.

What I Can Build Now (And What I Still Can't)

After 30 days, I am not a professional developer.

But I can:

• Build simple web apps that run in a browser
• Connect apps to APIs (like OpenAI, weather data, etc.)
• Save and load data using local storage or simple databases
• Debug errors by reading error messages and asking AI for help
• Customize existing code to fit my needs

What I still cannot do:

• Build mobile apps (that is a different skill set)
• Build anything that needs heavy security or handles payments
• Optimize code for performance
• Work on a team with other developers using version control

Basically, I can build tools for myself and close friends. I cannot build production software for thousands of users.

But for my needs, that is enough.

The Real ROI

The $4,200 I saved by building my own tools is great.

But the bigger win is that I am not stuck anymore.

Before this experiment, every time I had an idea for a tool or an app, I had two options: pay someone or give up.

Now I have a third option: build it myself in a weekend.

That is freedom.

I am not going to quit my job and become a software engineer. But I do not feel helpless around technology anymore.

When something breaks, I can usually fix it. When I need a custom tool, I can build it. When I have an idea, I can prototype it.

That shift in mindset is worth more than $4,200.

Would This Work for You?

I think it depends on two things.

1. Do you have a real problem to solve?

If you are learning to code because it sounds cool or because someone told you it is a valuable skill, you will probably quit.

But if you have a specific tool you need, a specific business problem, or a specific idea you want to test, AI-assisted learning works insanely well.

Motivation is everything. I stayed up until 1am debugging code multiple nights because I genuinely wanted the inventory tracker to work. I would not have done that for a practice exercise.

2. Are you okay with feeling dumb for a while?

The first week was humbling.

I did not understand anything. I was just following instructions like a robot. That is uncomfortable if you are used to being competent.

But if you can push through that phase and trust that understanding comes later, it works.

By week 2, things started making sense. By week 3, I was building confidently.

You just have to survive week 1.

The Bottom Line

A month ago, I thought coding was for people who went to college for computer science or spent years in bootcamps.

Now I know that is not true.

Coding is just problem-solving. And AI is the best problem-solving tutor that has ever existed.

It does not judge you. It does not get tired of your questions. It does not care if you ask the same thing five times.

If you have 30 days and a real project you want to build, you can learn enough to make it happen.

You will not become a professional developer. But you will stop being helpless.

And for most people, that is all you actually need.