The AI System That Built a 1.4M Audience

What if I told you you could build an audience of 1.4 million people in under 12 months? That’s what this episode’s guest has done using AI. Sabrina Romanov is one of the best AI people we’ve talked to, and she’s going to give you her entire AI playbook, including an end-to-end look at her workflow. She uses this workflow to take a single piece of content, turn it into many pieces of content, and turn all of that into millions and millions of people viewing that content each and every month. We’re going to break down her exact workflow step by step on this episode of Marketing Against the Grain.

Well, Sabrina, you are on a mission. You’re going to teach 10 million people how to use AI. And we’re going to teach thousands of people on this show at least — maybe hundreds of thousands. One of the things we wanted to talk about today was how you go from an idea to a basic vibe-coded app. But not just an app you play with and nobody ever sees or uses — an app you can actually distribute, get out into the world, and generate revenue, either for the app itself or for a related business.

I know you’ve spent a ton of time vibe coding. You’ve built a lot of cool apps, and you’re going to walk us through that. So I’d love to kick things off with a simple question: how do you get started? If somebody has an idea for an app, what should they do?

Sabrina:
Yeah, absolutely. I’ll walk through the entire process from start to finish. We’re going to build a small vibe-coded app — for example, an interactive quiz with an ROI calculator to help you capture leads for your business. Then I’ll show you how I use content repurposing automation to create one original high-quality piece of content and distribute it across eight other social platforms.

For me, that saves a ton of time because you make one original piece of content, you pour your heart into it, and then AI and automation intelligently repurpose it everywhere else in your own brand voice. So yeah, we’ll walk through that today. Happy to share my screen to kick things off.

Host:
Yeah, let’s go.

Sabrina:
Just to prove that this is a really valuable thing to vibe-code micro apps, this is a sample AI agents and automations directory that I vibe-coded about six months ago. In the past 90 days, it’s gotten 37,000 visitors — probably closer to 70,000 since inception.

The bounce rate is really high because it’s a single-page directory listing agents and automations. Let’s say you’re interested in content creation. You’ll see a lot of different automations here. For example, this AI agent automates social media carousels and slideshows. You can view the tutorial. You can download the template.

This is a really handy resource for folks who are brand new to agents and automations, don’t know where to get started, and don’t want to spend a lot of money building custom solutions yet. They want to tinker, download pre-built templates, and see how far they can go.

This little app has driven traffic to my bootstrapped website through subtle links — things like “get the best content creation and social media scheduling API.” That traffic flows back to my startup, which has gotten 1.3 million views just this year through organic social media content and vibe-coded micro apps like this.

This directory was built completely in Lovable. I vibe-coded it in just a few hours.

Today we’ll also walk through vibe coding a lead magnet in Lovable. For those not familiar, Lovable is one of many vibe-coding tools. I like it because it includes backend agents and AI — it handles your database, authentication, and things you’d traditionally have to wire up yourself.

So let’s try a prompt. We’ll use Lovable Cloud to create an interactive lead magnet quiz with a calculator. Let’s make it an AI readiness and ROI calculator to help clients understand how much time and money they could save with AI. We’ll ask 15 questions, generate a personalized report, and capture their email.

This initial prompt will take about five minutes to run because it’s scaffolding the entire codebase from scratch. It’s thinking about design, architecture, questions, scoring, and report layout. You can see it thinking through all of that now.

Host:
Sabrina, do you come from a technical background? And how hard is it for someone without a technical background to get started with vibe coding and build something robust and usable?

Sabrina:
My background is computer science and physics from Berkeley. I started an AI company right out of college in 2013 and sold it to Pegasus. So yes, I do have a technical background.

But I also run a women’s builder community, and many of them don’t have technical backgrounds and are successfully vibe coding. I think success depends on what you start with. If you try to build something massive on your first attempt, people usually fail. But something like a lead magnet is absolutely feasible today without technical experience.

Host:
Lead magnets are small apps you use to acquire traffic and then send that traffic to your main product or service. They’re related to what you sell, but they’re focused on one small, specific use case.

Sabrina:
Exactly.

Lovable is already setting up the backend. It asks for permission to control database storage. For lead magnets, you usually don’t need authentication, so we skip that. You can also add ChatGPT calls without managing API keys.

Lovable also gives risk assessments. It flags exposed keys — a common mistake with vibe coding — and can help you fix those risks automatically.

Host:
I love that. That’s been a big criticism of vibe-coding tools.

Sabrina:
You can see its reasoning as it works. I encourage people to read this. It’s modifying the database in real time, editing CSS files, and building the entire app from scratch. That’s why the first run takes a few minutes.

Host:
What’s the learning curve like? Does it get faster the more you do it?

Sabrina:
Most people are surprised how far they get the first time — maybe 70%. But the last 30% that makes the product unique takes most of the time.

The people who succeed build incrementally. They don’t start with a massive PRD. They start simple and add one thing at a time. They also read what the AI is doing, question it, and push back.

The people who struggle use AI to distance themselves from the process. They don’t want to understand what’s happening. That mindset rarely works.

Host:
That makes sense. Edits are where things get hard.

Sabrina:
Exactly. Start simple. My own app MVP was embarrassingly simple. You could rebuild it today in a few hours. But that was just the beginning.

Building your first version is the start of product development, not the end.

Host:
Okay, so now we have the AI readiness calculator.

Sabrina:
Right. It asks 15 questions, calculates a readiness score, estimates annual savings, and recommends next steps. You’d customize the questions and scoring for your niche, but in five minutes we went from nothing to a usable lead magnet that captures email and delivers value.

Host:
The best apps come from a strong point of view.

Sabrina:
Exactly. Prompts are a starting point, not the end state. You need to encode your expertise into the app so it delivers a meaningful before-and-after experience.

Host:
This replaces the old “PDF report” model. Code has become content.

Sabrina:
Yes. You can even personalize this further by researching the user’s company and tailoring the report. And email capture is built in — no extra tools required.

Host:
Now let’s talk marketing.

Sabrina:
The biggest misunderstanding is that AI fully automates high-quality content. I use AI as a sparring partner, not a replacement.

The lowest-hanging fruit is repurposing existing content. I grew Instagram from zero to 418,000 followers in a year by repurposing TikTok videos. TikTok is where I create original content; everything else is automated.

Another big lesson: AI is leverage. Skill times clarity equals leverage. The clearer you are and the more skill you have, the more you get out of AI.

Host:
So beginners shouldn’t start with automation?

Sabrina:
Exactly. Beginners should ignore AI at first and build the habit of creating content. Once consistency is there, then AI can help scale what’s already working.

Host:
That’s a really important distinction.

Sabrina:
Once you find a repeatable format that works, AI can replicate it across platforms. But you have to earn that first.

Host:
Why teaching? Why aim to teach 10 million people?

Sabrina:
Moving out of Silicon Valley showed me how little most people know about what AI can do. When my family visited and used Midjourney, it unlocked something creative and joyful for them. I realized I could teach this.

There’s a massive AI literacy gap. I want to help close it.

Host:
That’s a great mission.

Sabrina, this has been an incredible episode. You’ve shown how people with expertise can use AI to build products, grow audiences, and make an impact at scale. We’ll link your app, Substack, and everything below.

Thanks so much for coming on the show.

Sabrina:
Thank you. This was awesome.