Vibe Coding: How I Built an App During Lunch Break (And You Can Too)

By: Yessica Fontánez, Investment Analyst

Over the last year or so, the topic of AI for streamlining efficiency has been top of mind for everybody. At this point, AI capabilities are advancing on what seems to be an hour by hour basis, it’s no surprise that feats AI couldn’t accomplish just 6 months ago are possible now.

With this in mind, startup founders and individuals that are mastering AI tools like Claude, Gemini, and ChatGPT are gaining an indistinguishable competitive edge. Specially for early stage founders, these new tools are allowing small scrappy teams to accomplish more than what previous teams had with less — less capital, less time, less people.

These new AI savvy founders are leveraging these tools to focus on what’s most important for any business: gaining revenue.

However, prior to gaining revenue you need a product, a minimum viable product (MVP). You can launch an MVP in the market to start gaining those precious starting customers. For software startups, this is often the hardest part. If you aren’t lucky enough to pair up with technical founders OR be the technical founder yourself, making the MVP becomes the stopping point for many.

And sure, there are many work-arounds to this stopping point. Gaining the IP from the coding of the actual platform is still very expensive, time intensive, trial and error work.

So, here is where AI is swooping in to create a new fast-track to work; with vibe coding. If you haven’t yet been familiarized with the term, vibe coding is an emerging concept that leverages AI to code an app or website with a natural language prompt or prompts.

It’s something that I’ve been obsessed with during these past few weeks.

Before that, let’s review the fact that started it all for me. Recently, Y Combinator’s CEO said that AI wrote a whopping 95% of the code for a quarter of YC’s current startups. That’s actually insane, but if the startups in top startup accelerators are doing it, maybe it’s worth trying.

Thus, I tried it. I made a functional workout routine tracking app during my lunch break. I fully leveraged AI for the prompts, for the bug fixes and for the understanding of where things went. I’m happy to say that my exercise routine hasn’t been tracked like this ever in my life.

I have to give a mini disclaimer at this point that although I’m tech savvy and a wiz on an Excel sheet, I’m not at all versed in any coding language. Which means, if you’re a curious person at heart and willing to give things a try, you can create an app that you can later sell to customers.

So, if this has sparked some imagination in you, here is how I did it:

Step one: I found myself a vibe coding platform the old-fashioned way, through Google search.

You’ll be tempted to use ChatGPT to find the platforms, but at the moment of me writing this baby Gipi still didn’t know the term Vibe coding so it hallucinated a response. Good old Google still wins sometimes.

Step two: Pick one of the platforms that calls to you. I used Rork.

It provided 5 free prompts for the app creation which is great for testing. Remember when I said product use is cheaper now!

Step three: This was Gipi’s moment! Knowing my limits I asked ChatGPT to help me write my first Rork prompt.

After a couple of minutes answering all of GPT’s questions, I had a fully functional prompt with instructions on UI/UX, Front and back end, and features it should have.

Step four: Copy pasted GPT’s updated prompt to Rork and waited… quite a bit actually, so I went to lunch.

Step Five: I came back from lunch to an app! It was buggy and only semi-functional but it was mine!

FYI: Now, this next part is not my finest moment. I actually wasted the remaining free prompts trying to fix bugs one by one. You won’t do that, because you have hindsight. ❤

Step Six: I reviewed the agent’s error report and the locations it flagged. I copied the code to GPT and asked it to fix it. Then it was copy, paste and deploy. One bug fixed, now to the next. Until it was just right.

Step Seven: Then, I launched it on my phone, to test for breaking points and additional bugs.

Recommendation here: Press all the buttons at once. Don’t be nice to your app, you really want to test it. If it breaks, repeat step five.

Step Eight: Publish and use your brand new app!

There you have it, a super fast and simple way to build an MVP.

Am I confident it won’t crash at some point because of some unknown bug issue? No. But just by building it and talking about it I have my first test users. They’ve given me feedback on potential new features and have been using it to track their workout progress with me. Very nice for a lunch break!

Now, I don’t believe vibe coding will replace the need for software engineers in the near future. But it’s a great way for founders with non-tech backgrounds to test, iterate and deploy prototypes or features quickly and, importantly for the capital strapped, cheaply.

Founders who use AI for writing code will be able to conceive and deploy new features, receive instant customer feedback, re-iterate and re-deploy faster than their counterparts. Software developers that leverage AI will be able to find bugs and deploy fixes faster than their counterparts. Idea generation specialists soon to be multi-billion dollar exit founders will have probably used AI to generate the code to their startups.

And people like me who just wanted to track their gains will have an app on their phone just for them!

FUN FACT: As I was writing this blog post, Gemini launched a new agent to help developers code. So, I feel very validated.