Productside Stories

The Truth About Vibe Coding for Product Managers

Featured Guest:

Elena Avramenko, Founder and CEO | Modaal.dev
10/13/2026

Summary

In this episode of Productside Stories, Rina Alexin talks with Elena Avramenko, founder and CEO of Modaal.dev, about what vibe coding actually means for product managers, including the parts nobody talks about in the LinkedIn posts.

Elena’s story spans two chapters. The first: building PopAppFactory in Ukraine starting in 2009, a portfolio of more than 40 iPad apps, 6 million downloads, and an Apple App of the Year. The second: discovering Lovable as one of its earliest users, realizing what it unlocked for non-technical builders, and founding Modaal.dev to bring that same power to native mobile app development.

The conversation covers the hidden maintenance cost of AI-generated code, why Elena treats every agent like a junior developer and challenges every decision it makes, and why she believes craft will become the moat once AI-built apps flood the App Store. She walks through her team’s four-step workflow: vision, taste, debug, and delight, with her engineer now spending only about 5% of his time on what used to be exclusively engineering work. She shares the validation discipline she sees in her builder community, where most people kill five to seven ideas before committing to one. And she closes with her single piece of advice for product leaders trying to drive AI adoption: lead by example, build something yourself first, and invite your team to follow.

Key Takeaways

The hidden maintenance cost of AI-generated code.

  • AI writes code fast, but it never tells you what that code will cost to maintain. Elena explains why cheap code is not the same as good code, and why she treats every agent like a junior developer — question every decision, even when it comes from the best model available, because you have no idea what the long-term maintenance bill will look like until it arrives. Building a knowledge base within your AI tools through structure and specification can significantly reduce those costs over time.

Validate before you vibe code.

  • Your spare hours are an investment. Elena explains why most builders in her community kill five to seven ideas before committing to one, and why customer conversations should come before the first prompt. Treat your 10 free hours per week the way you would treat any other investment: with a hypothesis, a validation process, and an honest answer to the question of whether this is worth pursuing.

Vision, taste, debug, delight.

  • Elena walks through how her three-person team builds with AI. She builds an ugly working prototype (vision), her designer replaces the AI-generated UI with real craft (taste), the whole team debugs different user flows (debug), and they add micro-animations and personality together (delight). Her engineer spends about 5% of his time on the work. That ratio is the signal.

Craft becomes the moat.

  • When anyone can ship an app, human-crafted design is what earns the revenue. AI-generated UI apps can validate a hypothesis, but if you want a real revenue stream you need to invest in human craft. Once AI-built apps flood the market, the ones that top the App Store will be the ones where someone cared about the details AI never notices.

Leaders go first.

  • Want your team building with AI? Build something yourself first. Elena’s clearest piece of advice from her MBA: if you want change in your organization, lead by example. Show up to build-a-thons, build a hobby project, share it with your team, and let them follow. Telling people to use AI without doing it yourself is the fastest way to make sure nobody does. The shift toward generalist skills is accelerating in the AI era and that is good news for product people.

Chapters

  • 00:00:00 – Introduction to Elena Avramenko and Modaal.dev
  • 00:01:07 – From iPad Apps to AI: Elena Avramenko’s Path to Modaal.dev
  • 00:05:26 – When Building Gets Cheap, Does Product Judgment Disappear?
  • 00:09:45 – The Hidden Maintenance Cost of AI-Generated Code
  • 00:11:00 – Different Archetypes of Vibe Coders
  • 00:14:55 – Validate Before You Vibe Code: Treat Your Time Like an Investment
  • 00:18:25 – Bring Your Own Agent: Where Vibe Coding Tools Are Heading
  • 00:20:56 – Vision, Taste, Debug, Delight: How a Three-Person AI Team Builds
  • 00:27:48 – What Vibe Coding Means for Engineers, Generalists, and Careers
  • 00:30:57 – The Rise of Solopreneurs and Generalists
  • 00:36:30 – How Product Leaders Can Drive AI Adoption: Lead by Example
  • 00:38:59 – Where to Connect and Closing Remarks

Why Listen to This Episode?

In this episode, you will:

  • Learn how to approach AI prototyping for non-engineers without skipping product discovery
  • Hear why human craft will become the moat once AI-built apps flood the App Store
  • Find out what “bring your own agent” means for vibe coding tools and your credit card bill
  • Get Elena’s rule for leading AI adoption on your team: build something yourself first, then invite your team to follow

If you’ve ever wondered whether product managers should vibe code, this episode gives you the honest answer, maintenance bill included.

Introduction to Elena Avramenko and Modaal.dev

Rina Alexin | 00:00:01 – 00:01:07

Hi everyone, and welcome to Productside Stories, the podcast where we dig into the very real and raw lessons learned from product leaders and thinkers all over the world. I’m your host, Rina Alexin, CEO of Productside, and today I’m talking with Elena Avramenko, founder and CEO of Modaal.dev, where she’s helping non-engineers build real native mobile apps with AI.

After over 15 years across companies like Miro, Lego, and Microsoft, and a first startup, PopAppFactory, that shipped over 100 apps and won Apple’s App of the Year, we’re going to be talking about vibe coding and prototyping with AI, how product intent turns into working prototypes very fast, how teams are experimenting with far less engineering cost, and how all of this is moving the bar for what counts as an MVP. Welcome, Elena.

Elena Avramenko | 00:00:59

Thank you, Rina, for such a wonderful introduction. Happy to be here and share the story.

From iPad Apps to AI: Elena Avramenko’s Path to Modaal.dev

Rina Alexin | 00:01:07 – 00:01:20

Tell us a bit about your own story. How did you end up building Modaal.dev?

Elena Avramenko | 00:01:20 – 00:05:26

Building Modaal actually has two sides to the story.

The first side began in Ukraine in 2009, when mobile was just getting started. My university friend came to me while I was working at Microsoft. He said, “Mobile is going to be huge. Join me. I’m opening a mobile agency.” I’m very entrepreneurial, someone who likes to jump into unpredictable situations and untested waters. So we started together, initially as a mobile agency building products for clients. But very fast we realized it was more interesting to build a product company.

Around that time I became a mom, and I understood that kids’ educational products for iPads specifically was something I really wanted to build. We built our first app. Apple reached out and said, “You’re building an incredible visual style and fantastic interaction every time. Just message us when you release something. We’ll feature you.”

That was the beginning of PopAppFactory, our indie lifestyle business in Ukraine:

  • More than 40 apps
  • More than 6 million total downloads
  • 1 million in total revenue, all organic, driven by Apple featuring

The second part of the story started last year. In February or March, I discovered Lovable and became one of its early users. I was completely impressed. As a non-technical person, this was something I had wanted my whole life. I immediately understood that it was huge for people with ideas, taste, and entrepreneurial ambition.

I built several web projects and participated in different hackathons on Lovable. And I understood: I want to build something for vibe coding for non-technical people like me. Specifically for solopreneurs and indie builders, could I build a tool that helps them release production-ready apps for revenue with little or no developer help?

I spoke with my co-founder from PopAppFactory. He was at Meta. We said let’s do something. We released in February. Right now we have 300 users, a fully bootstrapped company with three people. Product people, by the way, are my most active and most interesting users. Modaal feels like my destiny, because I’m an entrepreneur by heart and I’m very keen to help people like me build solo businesses with AI and proper tooling.

When Building Gets Cheap, Does Product Judgment Disappear?

Rina Alexin | 00:05:26 – 00:09:45

You’re making it accessible for everyone to say “I have an app for that.” But I think about what happened when building was still expensive: engineering at least forced some judgment. Back in 2015 or 2016, I worked with an organization where a leader told the team, “I want an app for this.” And what this actually was: a single screen with just their customer support number. Just a phone number. Now that building is cheap and more accessible, could that actually be a problem?

Elena Avramenko | 00:06:29 – 00:09:45

I think it is genuinely a huge problem. Before AI, cheap meant bad code. With AI, cheap means code with unpredictable quality. You have no idea whether what you built is good code or bad code. It’s just a piece of software built from a general knowledge base, and you have no idea what the maintenance cost will look like.

Before, if I worked with a freelancer, the trade-off was obvious. I could select someone with a good rating, who had protocols, who could be held accountable for bugs. With AI, I build something and I have no idea:

  • What is the maintenance cost?
  • What will it cost to scale this?
  • What will it cost to make this work for profit?

If you’re building for hobby, you can tolerate this. It’s an exploration. But if you’re building for profit, maintenance is a huge question mark that most beginner vibe coders do not think about.

“Before AI, cheap meant bad code. With AI, cheap means code with unpredictable quality. I have no idea what the maintenance cost will be.”

My personal discipline is to challenge everything that comes from the agent. Am I sure this decision is right? I dig deeper. I treat the agent like a junior developer: even with the best model, you need to challenge every decision and ask whether it is actually right. By doing that consistently, you can reduce future maintenance cost, but the problem doesn’t go away entirely. Many people, especially beginner vibe coders, simply don’t understand this.

The Hidden Maintenance Cost of AI-Generated Code

Rina Alexin | 00:09:45 – 00:11:00

If your vision for Modaal is to help people create apps that will actually drive profit and revenue, how are you addressing this problem for your users?

Elena Avramenko | 00:10:02 – 00:11:00

This is exactly why we built Modaal. I started building my apps with Claude Code, just Claude Code. Claude Code works fantastically for three- or four-page apps. Your phone-number example: you don’t need Modaal for that. Claude Code will do it and you pay nothing except your subscription.

What Modaal addresses is the hidden maintenance cost. We give every project a knowledge base and an app structure. We use any available model. The long-term maintenance is reduced by giving a lot of structure and guidance, injected from my co-founder’s 20 years of experience building enterprise-level software.

You will still have some maintenance cost. And I genuinely believe you still need developers to build a successful product. But with Modaal, at least we can decrease that cost dramatically compared to going directly into Claude Code without structure.

Different Archetypes of Vibe Coders

Rina Alexin | 00:11:00 – 00:11:36

What about when people have an idea but no real product management knowledge, no understanding of the market, and they build something nobody needs? Is there a tell for you that someone is about to build something wrong?

Elena Avramenko | 00:11:36 – 00:14:55

I can give different examples from my personal community. We have a few distinct archetypes:

  • People who have never developed any software. This is the most difficult category. They have no understanding of user flows, no concept of a PRD. They have an idea, but no framework to evaluate it. Sometimes they have a distribution channel, which changes things completely. For example, I have a doctor in my community who wrote a PhD on habit formation and already has a network of practitioners she can sell to immediately. For her, an app is a real business opportunity worth pursuing. But the learning curve to build it herself is steep.
  • People who have already built software and understand workflows. These people are typically working full time and building something outside their job, on a path to becoming an entrepreneur one day. They vibe code during weekends. We see huge spikes in prompts sent to Modaal on weekends, because people are genuinely trying to build something that will eventually give them enough income to stop working full time.
  • Product people. For them, this is the most natural environment. The whole approach is peer-to-peer and driven by specification. They build in a comfortable, structured way because they already think in terms of user flows, hypotheses, and outcomes.

Vibe coding is a journey and everyone starts from a different point. The most important thing is intent. If you want to learn, embrace this new reality, and you have time and dedication, you will make it happen even without a product or engineering background. Claude Anthropic Academy, for example, is a fantastic free resource to start with.

Validate Before You Vibe Code: Treat Your Time Like an Investment

Rina Alexin | 00:14:55 – 00:15:10

Is it ever too early to start vibe coding?

Elena Avramenko | 00:15:15 – 00:18:25

Let me split this into two parts.

Is it too early to try vibe coding in general? No. The faster you get to a vibe coding tool to try something, the better. But do not over-invest without validation.

I see people at build-a-thons investing hours into an idea that came to them 10 minutes before the event, without any validation whatsoever. That’s fine as a way to test the water and see if vibe coding is for you. But if you really want to build an idea you believe in, approach it like an investment project:

  • You have maybe 10 free hours per week outside of work
  • You can spend those 10 hours without any plan, and both your time and your credits are expensive in their own way
  • Or you can treat it like an investment: understand the problem, research the existing solutions, speak with potential customers, and validate whether it is actually worth pursuing before you write a single prompt

“Usually people are killing five, six, seven ideas before they start investing into something they really want to build long-term.”

That discipline is what separates the builders who eventually ship something meaningful from those who burn through time and credits chasing the wrong thing. The original idea is almost never the one that sticks. Even Brex, the company known for corporate credit cards, started as an AR hardware startup. You really need to get comfortable letting go of your ideas and your ego, because the pivot is usually the point.

Bring Your Own Agent: Where Vibe Coding Tools Are Heading

Rina Alexin | 00:18:25 – 00:18:48

I’d love to hear about your own process. Can you walk us through what it looks like when you, your designer, and your engineer are each building with AI?

Elena Avramenko | 00:19:09 – 00:20:56

I’d love to show you. The best way to explain it is to show how it actually works in Modaal.

What you see is Modaal as a Mac application. There is a setup needed to get started, roughly 10 minutes. One important thing I want to share: the business model for vibe coding tools is shifting. Initially, tools charged per credit. The model that is now dominating, and I think is better for users, is called bring your own agent:

  • You use the vibe coding tool as a workflow layer and plug in your own model subscription
  • I use Claude. I plug in my Claude subscription and I don’t pay per credit. I pay per seat and I build whatever I want.
  • That is more predictable from a cost perspective
  • It is what the builder community has been asking for, because unpredictable credit spend is one of the biggest complaints

The app I’ll show you is Memory Lane, a family digital memory box you share with close family members and close friends. You can post memories in your lane, share with someone in your close network, and see shared memories with reactions. It’s not a social feed. It’s a close network. And this is where I’ll walk you through the four-step workflow my team uses.

Vision, Taste, Debug, Delight: How a Three-Person AI Team Builds

Elena Avramenko | 00:20:56 – 00:27:48

Our workflow is built on four steps: vision, taste, debug, and delight. I own vision. My product designer Vika owns taste. Our engineer Ivan and I share debug. All three of us collaborate on delight.

  • Vision (Elena’s part). I am responsible for validating the idea, understanding what we are building, articulating the hypothesis, and defining the ICP. All of those details come before building. Then I embed this vision into what I call an ugly functioning prototype. I go into Modaal first and build a working app, but very ugly: AI-generated design, functioning but not delightful. Some flows may break. But the general idea is clear enough to hand off.
  • Taste (Vika’s part). Once the ugly prototype is built, my designer takes over. She is responsible for replacing the AI-generated design with something well-crafted. My true belief is that in the future, as the App Store fills with vibe-coded apps, craft will become the moat. Human-crafted apps will be the top revenue-generating apps. Yes, you can release an AI-generated UI app to validate a hypothesis. But if you want to generate real revenue, you need to invest in human-crafted design. Vika creates mood boards, brings ideas from Figma and Claude Design, and builds the design layer on top of what I’ve already built functionally.
  • Debug (all three of us). We each go through different user flows. I debug my lane, Vika debugs the profile, and Ivan handles more complex flows that neither of us can debug alone. When I debug something, I start with a new chat and use plan moat. Plan moat treats the issue like an epic. I write what needs to be debugged, tag the relevant specification, provide user stories of what is broken, and the agent adds a debugging specification linked to the original. Vika can then see exactly what I fixed, and I can see what she fixed. The whole specification is shared and transparent for the team.
  • Delight (Elena and Vika together). We add the micro-animations, the funny phrases, the little details that make an app feel emotional and memorable. A hand-drawn illustration here, a micro-animation when you add a memory there. These are the things Vika suggests and we decide on together, then we vibe code each one separately.

“Craft will become the moat. Human-crafted apps will be top revenue-generating apps.”

Ivan’s time split: roughly 5% of his time. The rest is divided more or less equally between me and Vika. That ratio is not accidental. It is the point.

Ivan has also been building the Android version. Modaal now supports Android native Kotlin as well, which means the whole team has visibility into who is doing what through the Git repository connection at any time.

What Vibe Coding Means for Engineers, Generalists, and Careers

Rina Alexin | 00:27:48 – 00:30:57

Your engineer is spending only about 5% of his time on work that used to be entirely an engineering role. What does he think about where engineering is going over the next few years?

Elena Avramenko | 00:29:00 – 00:30:57

My co-founder is pessimistic about engineering in general, not because he doesn’t believe in engineers, but because of what he observes in the market. What we are seeing:

  • Large companies have reduced headcounts because AI productivity has increased. Smaller teams can do what larger teams used to do.
  • Senior and staff-level engineers are relatively safe. They can always find a job and can always find founders who understand the value of a good engineer who gives guardrails to LLMs and builds structure that reduces long-term maintenance.
  • Young engineers are struggling the most. They are being substituted by LLMs. That creates a real problem: if we don’t invest in growing junior engineers, where do the senior engineers of the future come from?

But from another side, there was probably over-investment in engineering in general. Everyone was told to become an engineer. Now what I think we need more of is product people and designers.

And engineers who are adaptable are finding entirely new things to do. My co-founder is now doing user interviews, which he never did before. He is starting to record TikTok videos to show other engineers what is possible. Engineers who say “I’m not going to learn any new skill, I’m not going to do prototyping or user interviews” will struggle. Engineers who adapt will be quite safe.

The Rise of Solopreneurs and Generalists

Rina Alexin | 00:30:57 – 00:36:30

You’re building something that lets people work on a side hustle they couldn’t have pursued before. What do you think happens more broadly to careers and to people now that they are able to build?

Elena Avramenko | 00:32:42 – 00:36:30

A lot is happening. My observation comes from more than 40 build-a-thons across Lovable, Modaal, and other communities, mainly in Europe. What I observe: some people genuinely love a corporate career. Predictability, a career ladder, stability. That’s completely fine. But there is a subset of people within corporate environments who are not enjoying their lives there. They have an entrepreneurial gene. They have ideas. They don’t fit the corporate ladder structure.

Before AI and vibe coding, pursuing that entrepreneurial path required a lot: saving money to hire developers, quitting your job, making a large financial commitment before you had any evidence that your idea would work. That barrier was real and it stopped a lot of people.

Now it is genuinely possible to step by step build a solopreneur career while still fully employed, and then finally make the leap. I see it in my community. People experimenting during weekends. One woman I know is building an app to help with fermenting and marinating, a genuine science in Eastern European cooking with different temperatures, ingredients, and timing for every jar. That’s a niche. But if she can monetize it, she will leave the corporate job she doesn’t love.

“The most difficult thing is to be successful in what you don’t like. Especially at a certain point in life when you are 15 years in, you’ve gone through the career ladder, and you can’t stop because the salary is good and you’re an expert, but you don’t enjoy the actual work you do.”

Vibe coding is a genuine unlock for those people. From the company side, it is also raising the value of generalists. Five or six years ago I would describe myself as a generalist and struggle to explain my value. Society rewarded extreme specialists. Now I see even large companies actively wanting generalists, people who can cover multiple roles. Some roles are already combining. The teams are getting smaller. AI is bringing that change and I think it is mostly positive, even if the transition is difficult for some.

How Product Leaders Can Drive AI Adoption: Lead by Example

Rina Alexin | 00:36:30 – 00:38:43

Elena, you’ve spoken like a true entrepreneur. For product people and product leaders who are still trying to figure out where vibe coding fits, and who have teams and organizations that are still learning to use agents and AI tools, what advice would you give to those trying to empower their teams to do more?

Elena Avramenko | 00:38:43 – 00:40:43

One thing I took from my MBA that I believe is a permanent truth from organizational behavior: if you want change in your work, you need to lead by example.

I have been to a lot of build-a-thons and hackathons. The pattern that works is heads of product and VPs of product coming to those events, building themselves, bringing their teams, and running internal team build-a-thons. They give the signal: “I’m building myself. I have a hobby project. I built this thing for my family.”

The practical starting point:

  • Take any idea you want, take any tool you want
  • Build something: no security concerns, no business stakes, just vibing and having fun with it
  • Run a lunch and learn where everyone shows what they built

“Lead by example. Show that you are a builder. You built this thing for your family or your hobby. Show the tools. Inspire people. That is the most important thing.”

The pattern that fails: individual contributors coming to build-a-thons and saying their manager told them, “Why did you build this without AI? Next time you should be using AI.” That is a show-not-tell failure. As a leader, you need to show, not tell. Be a practitioner yourself. Inspire your team by doing it, not by directing it.

 “If you want change in your work, you need to lead by example. Be a practitioner yourself and inspire your team to follow you.”

How to Follow Elena and Closing Remarks

Rina Alexin | 00:40:54

Leaders, you have to lead by example. Well said, Elena. Thank you so much for joining me today. How can our listeners follow you after they listen to this episode?

Elena Avramenko | 00:41:05

I’m on LinkedIn. Message me there. I’m happy to reply to everything I receive except the AI-generated messages, so please write me a personal message. I’m happy to help, happy to see you in my community of mobile builders in the Modaal Discord. But LinkedIn is the best place.

Rina Alexin | 00:41:26 – 00:41:44

Wonderful. And thank you all for listening to this episode of Productside Stories. If you liked today’s conversation, please don’t keep it to yourself. Make sure to subscribe so you don’t miss a future episode. I’m Rina Alexin, and from all of us at Productside: let’s do product better, together.