I Got Tired of Writing Boilerplate — So I Let AI Do It For Me

I’m done wasting time on boilerplate and basic styles.
So, I made AI handle it for me. Results, thoughts, and revelations — inside.

TL;DR in Two Sentences

I try to cut down everything in development that can be automated, so I still have energy for creativity and complex mechanics. For that, I use AI. I’m not afraid it will replace me — instead, I’m learning what it can do well on the second try, what it tends to mess up, and what’s still faster to do by hand.

Let’s vibe-code, folks. 🚀

Why I Did This

Developers are lazy. Especially when it comes to the first steps: setting up a project, configuring the environment, pulling in all the libraries.

That’s why we invented starters, frameworks, templates — dragging them from one project to another. And still, we often waste tons of time tweaking configs for a new task. By the time everything is ready, honestly, I already feel lazy to start building the actual thing (at least speaking for myself).

So, when I had to migrate a React app to Vue — and not just Vue, but Nuxt — for my students, I sighed deeply. And then I thought: Nope, I’m not doing this by hand. I’ll let AI handle configs and boilerplate.

The stakes were low — this project was just a sandbox for my mentees, and I already had enough real work. No risk, pure fun. Agree?

(Note for non-techies: the task was to move a project from one framework to another. They serve a similar purpose, but with syntax and structural differences, which means a lot of code rewriting.)

Tools I Used

Cursor AI
Just getting started with it. It promises to replace my IDE and keep me from running around different tools. We’ll see. So far, I like how it gives inline suggestions while coding. A colleague actually nudged me to try it.

ChatGPT Plus
I don’t just code here — I use it for brainstorming, briefs, metaphors, and idea generation. Sometimes I even ask for reference illustrations or creative sparks for pet projects. GPT is great for switching tones and following prompts, but… it still hallucinates links and case studies way too often.

Perplexity
That’s where I fact-check GPT. If GPT gives me a case, I drop it into Perplexity to confirm it’s real (spoiler: too often it’s not). It cites sources, compares, and points me to related materials. Think of it as a mix between a fact-checker and a search engine.

How It Went

It started with Cursor AI and some manual setup — GitHub auth, repo configs, export settings. I created a blank Nuxt project myself (didn’t want to get too cocky) and then gave Cursor my first prompt:

This will be a pet project — a music player. It will have login and signup pages (almost identical, just different fields), a categories page for playlists, a “My Playlist” page, a main page with all songs, and a “Not Found” page.

In other words: the basics — routes, structure, styles. Cursor delivered. Nice, but not perfect. By default, it threw everything into Tailwind. I’m not a fan.

So I asked:

Please use scoped styles instead of Tailwind.

Result: a clean skeleton project in about five minutes.


User Interface:

Not bad. I kept feeding prompts while half-watching Telegram and videos. Code vibes.

Checking for Real Use Cases

While Cursor was working, I got curious — what cool real-world projects exist with Cursor AI?

ChatGPT led me to a Reddit thread where someone asked the same question: “Show me real apps, not just pet projects.”

One user shared an app in production with thousands of users, saying:

“It’s a mistake to think these tools are only good for boilerplate and toy projects… They already write most of my production code.”

Mind-blowing.

I followed a link to an AI app builder — but it cost $30 for 10 credits, so I passed.

Another user shared a free app built almost entirely with Cursor, tweaking just a few parts manually. That gave me confidence to keep pushing forward.

Back to My Project

Skipping the UI for now, I went straight into global state management and API calls. Prompts in, quick fixes when I forgot to define data structures, a little manual tweaking — but all fast and painless.

Then came Pinia integration. Cursor added the code into the project, but missed installing the package. I fixed it, and it worked like I had done everything by hand.

End result: a functioning app skeleton in about 30 minutes.

Sure, there were rough edges — for example, every track shared the same isPlaying state (so if one song played, all looked like they were playing). And I still had to polish styles, fonts, and small quirks. But hey, we skipped the most boring part of development.

The Takeaway

With AI, I had an app 80% ready in half an hour. The last 20% — well, you’ll still need to get your hands dirty.

But here’s the kicker: I only spent maybe 6 minutes actively prompting. The rest of the time? Reviewing and vibing.

And honestly, reviewing code feels way better than grinding through boilerplate you’ve written a hundred times before.

Why?
Because your eyes stay fresher, you spot risks earlier, and you think more about scalability and design. You get to think instead of type.

Even with this simple app, it felt great.