I Didn’t Know What Vibe Coding Was Until I Did It
I Never Really Cared About “Vibe Coding”
2025 was the year people suddenly started talking about “vibe coding.”
I had heard the term before, but honestly, I never really searched for it or looked into it that much.
Mostly because I was too lazy.
But I always liked the idea of making apps.
I think it’s just permanently built into my DNA at this point.
Why I Made a Lunch Budget App
My first vibe coding app was a lunch budget tracking app for office workers called “Bob.”
Before, my company used to provide meals, so I never had to think much about lunch expenses.
But after the company moved offices, the system changed.
Instead of free meals, the company started giving us a fixed monthly food budget on a card.
And we had to survive on that amount for the entire month.
Once the month ended, the balance reset.
Which basically meant:
If you ate expensive meals early in the month, you either had to spend your own money later…
or survive on cheap meals near the end.
And honestly, the amount they gave us did NOT match Gangnam prices at all.
If you didn’t carefully manage your spending, there was a very high chance you’d end up using your own money before the month ended.
Sure, the card app already showed how much money was left.
But I kept thinking:
“Wouldn’t it be nice if there was an app that calculated how much I could safely spend for the remaining workdays?”
And also:
“Maybe other office workers need this too?”
And, well…
“If people actually use this and I somehow make money from it… that would be pretty nice too.”
I mean… making money is always nice.
So I started building the app.
Wait… This Is Actually Vibe Coding?
At the time, I didn’t even know what “vibe coding” really meant.
But looking back now…
Yeah.
That was definitely vibe coding.
I first organized the systems I needed.
Although honestly, there wasn’t much to organize.
The app just calculates money.
Then I broke features into smaller parts and asked ChatGPT to build them one by one.
If I said:
“Subtract the amount when I enter money.”
GPT handled it pretty well.
But the moment I said:
“Make it clickable on a calendar UI AND track expenses by date.”
…it started struggling really hard.
HEY!
YOU IDIOT!
That’s when I realized:
My Cannon — my GPT assistant — was actually kind of dumb.
Cannon! You idiot!!
My Cannon Was Dumber Than I Expected
Even though the app only had a single main page, it still took around a month to build.
Technically, I could keep developing almost 24 hours a day even on the free version, because once I ran out of usage, GPT would simply switch to a weaker model and continue answering anyway.
But there was a huge problem.
If I didn’t fully understand the code Cannon generated, I couldn’t properly explain bugs or issues back to it.
And Cannon also didn’t fully understand what I wanted.
So debugging took FOREVER.
Honestly, if I include all the trial and error, it probably took almost six months.
But now things are different.
The current GPT models are much better, and my prompting skills are way better too.
If I rebuilt the same app today, I honestly think I could finish it within a week.
The Hardest Part Wasn’t Coding
Coding wasn’t actually the hardest part.
The hardest part was all the legal and release-related stuff.
Privacy policies.
Ads.
Review processes.
Required documents.
That stuff terrified me because I knew absolutely nothing about it.
And surprisingly…
That’s where Cannon helped me the most.
For the review process, I actually paid a professional service for help.
Google Play required a 14-day testing period, but almost everyone around me used iPhones.
And since I developed the app in Android Studio without owning an iPhone myself, I only released it on Android first.
Paying someone definitely cost money.
But honestly?
It was incredibly convenient.
I barely had to worry about anything.
During the testing period, I updated the app, fixed bugs, and applied feedback from testers.
Since it was my first release, having help from experienced people made a huge difference.
Releasing My First App Felt Amazing
After the testing period ended, I was finally able to release the app.
And because my first release process went surprisingly smoothly…
I genuinely thought app releases were always this easy.
(I was VERY wrong.)
I installed my app on my parents’ phones first.
Then I proudly showed it to some coworkers.
Since we all share the same lunch budget system at work, some people actually started using it consistently.
Even now, one of my managers still uses it and gives me feedback.
I also use the app myself almost every day.
After all, I built it because I genuinely needed it.
The Market Is Tiny. But I Still Love It.
The problem is…
This market is EXTREMELY niche.
Right now, I think the only active users are me and my boss.
So I constantly wonder:
“How do I improve this app?”
“How do I promote it?”
“Can this actually grow?”
But honestly…
Even with the tiny user base, I still love it.
Because it’s MY app.
And it solves MY problem.
If you happen to work at a company with a similar lunch budget system, maybe give it a try someday.
https://play.google.com/store/apps/details?id=com.bab.lunchbudgettracker
Now I’m Building More Apps
Right now, I’m already working on more apps.
1. A Plant Watering Reminder App
2. A Travel Expense Tracker
And this time…
I’ve been getting a lot more help from Claude instead of GPT.
Sorry GPT.
But for coding?
Claude is waaay better right now.
For the plant watering app, I tried getting real testers this time.
But apparently asking people to keep an app installed and actively test it for 14 days is harder than I expected.
So my first review attempt failed.
I’ll probably need to gather testers again and retry the process.
As for the travel expense app, it’s still in the early planning phase.
But I really want to release it before June ends.
Because I’m planning to travel to France in July.
I want to use the app myself during the trip, gather real feedback, and then write about the experience on this blog afterward.
I Still Don’t Fully Understand Vibe Coding
Honestly, I jumped into vibe coding without even properly understanding what vibe coding was.
But somehow…
I’ve been enjoying the process a lot.
I just wish Claude gave me more tokens.
And I hope more people eventually use my apps and send me feedback.
And hopefully…
One day…
My Cannon becomes a little smarter too.
댓글
댓글 쓰기