When AI writes 100K lines of code, QA becomes the whole job

Artur Balabanskyy runs Tapforce, an AI-first development agency, and he has a problem that nobody in the industry talks about honestly. When AI can generate 100,000 lines of code in a few hours, you don’t stop having a 100,000-line problem. You just have it faster. The bottleneck in software development hasn’t gone away. It’s moved, from writing code to validating it, and most agencies haven’t caught up to what that means operationally.

Balabanskyy has been coding since he was 10, has built iOS products through the Flash-to-mobile transition, and has run traditional development teams before AI tools changed the calculus entirely.By 2023, he was watching ChatGPT produce buggy but functional code and saw the shift coming. “I could build an MVP in a few minutes and give it to my coders to complete,” he tells The New Stack. “It wasn’t great code, but it was code.” What followed that observation is where the real story is.

AI grandma

Artur Balabanskyy did not take a straight path to a place where he could truly embrace AI development. His path bent around other people’s expectations, then came back to something he truly loved.

Balabanskyy grew up in Lviv and started coding at ten. He turned that love of computers into a multi-million dollar business. And it all started because he wanted to recreate his grandma with AI.

Artur’s story started in the late 1990s.

“A kid introduced me to Delphi and Objective Pascal,” he says. “I loved it.” That early pull toward building never left. Even then, he was not just learning syntax. Instead, he was trying to recreate the world around him in code. 

“When I was 13, we were building an AI clone of my grandma with if-else statements,” he says. “I thought I could keep her around forever if I made a copy of her on my computer. I didn’t realize how hard it would be just to communicate with something that was basically a very early version of machine learning.”

“When I was 13, we were building an AI clone of my grandma with if-else statements. I thought I could keep her around forever if I made a copy of her on my computer.”

Instead of going straight into engineering, Balabanskyy pursued international economics. He graduated with honors from a Master’s program, focusing on global markets and economic systems. He then entered a PhD program, on track for an academic career.

He chose to walk away.

“I realized I didn’t want to stay in theory,” he says. “I wanted to build things that have immediate impact.”

That decision changed the trajectory of his career. It also shaped how he approaches technology. He does not see code as an isolated craft. He sees it as a tool that operates inside larger systems, markets, incentives, and constraints. That perspective is part of why companies trust him with decisions that go beyond implementation and into strategy.

“I was like, ‘Ugh, it was not my thing,’” he says. “I was stuck doing things where I couldn’t see actual results. I wasn’t building as much as reorganizing. I was taking things people thought or said and creating new ideas, but they didn’t do anything. It didn’t have the power of code.”

While studying, he returned to code through an IT school. He was happy to start working on less theory and more practical problems. He was spending long hours building instead of reading and his love of coding started to supersede his love of economics. After three years, he walked away from the PhD.

From there, things moved quickly. He began teaching. Then a connection pulled him into a small development team. His early work was in Flash. That world disappeared fast.

“I still remember when Steve Jobs rejected Flash on iOS and we had to port our application to mobile. It was frustrating because everyone knew Flash so well but learning to code for iOS changed my life.”

That pivot made him an iOS engineer. He spent years building products, teaching others, and working with teams like Daily Steals. At the same time, he kept solving problems. He built tools for his wife’s work in the courts. He was trying to cut down repetitive paperwork so he built early expert systems that could go through detailed documents and bring out pertinent information.

“I was trying to solve and improve other people’s work and speed and productivity,” he says. “The goal was the same as it was when I was building my AI grandma: I wanted to capture and manage reality using code. It was so hard back then. The tools weren’t right and everything was an if-then statement, a fact that was super frustrating to me. And this wasn’t even that long ago. I started coding for iOS in 2010 and back then it was like pulling teeth.”

By 2017, he shifted into management with less coding and more coordination. He dealt with teams, clients, and project scopes. He thought his job would remain the same for the next decade. He was wrong.

“By 2023 it was clear something was happening in the programming world. ChatGPT had launched a year before and people were using it to produce code that was buggy and nearly useless,” he says. “But they were producing code. At first I was scared, but I saw something emergent in the way these tools were being used. I was stuck behind a spreadsheet making sure that each aspect of a project was being handled by coders who were just doing a job. By using a tool like ChatGPT I could build an MVP in a few minutes and give it to my coders to complete. It wasn’t great code, but it was code.”

Almost a decade later, that decision has turned into something tangible. His agency, Tapforce, has helped build mulit-million dollar products while serving dozens of clients across a mix of product and platform work. It is no longer a small development shop taking on one-off projects. It is a business with repeat customers, predictable income, and a clear sense of direction. The shift toward an AI-driven model is not a gamble for him. It is a continuation of the same instinct that pulled him back to coding in the first place, a belief that the tools will change, but the advantage stays with the people who know how to build.

His career in development has paid off in a direct way. Years spent writing code, shipping products, and working through real constraints have given him leverage that is hard to fake. He understands how systems break, how clients think, and how to turn an idea into something that runs. That experience now compounds. As the agency moves deeper into AI, he is not starting from zero. He is applying a decade of practical knowledge to a new layer of tooling, and that is what is driving growth.

“As you know, I’m staying in the middle. The same way I’m excited, the same way I’m stressed.”

The excitement is obvious. AI tools collapse time and work that once required teams now fits into a single workflow.

“I can spend a couple of days building something that would require a couple of months before because I know where to guide the AI. I don’t consider it scary. Instead, I consider it freeing. I can build quickly, iterate, and then deliver a final product that works.”

Balabanskyy doesn’t think that AI will destroy dev work. It will change it.

“AI makes anyone a developer. That’s bad for dev shops that aren’t ready. Most dev shop managers think the money is in the coding, the long hours it takes to get from zero to one. That’s all been flattened. So what’s left? A CEO who vibe codes an app that exposes her entire database who needs help building something production ready. A solo founder who needs a partner to tell him whether a piece of code is actually working or not. A government employee who can use AI only up to a point.”

“AI makes the easy stuff easier. But there’s still plenty of room for architects, creators, entrepreneurs. If the computer was a bicycle for the mind, AI is a jet. But just like you need to know how to ride a bike or pilot a jet, you need to know how to build.”

“If the computer was a bicycle for the mind, AI is a jet. But just like you need to know how to ride a bike or pilot a jet, you need to know how to build.”

Balabanskyy says that AI changes the structure of a company. It takes fewer people to perform fewer roles and there is less separation between disciplines. But he thinks that makes it easy for coders to do what they really love: dream.

“With AI you are not just a backend engineer. You can do backend and frontend, and you can even create massive systems. You turn into a project manager, backend dev, frontend designer, QA expert. And because it’s so quick and easy you can build instantly, resulting in less cost and fewer headaches.”

He sees the agency model itself under pressure. Clients are starting to believe they can do the work on their own. That is the threat. But it is also the opening. Instead of selling labor, the shift is toward building products, tools, and systems that scale. He thinks that coders become something else entirely.

“Maybe they’re agent operators. Maybe they’re project managers. You still need expertise in these roles and things will change constantly until we decide how humans fit into this loop. It’s like we’re in a dust storm and we’re waiting for the sand to clear.”

Still, he is clear on one thing: his role is changing from writing code to directing systems that write code, and then checking them.

“Now we need to QA even more. When you can spit out 100,000 lines of code in a few hours, you have a huge problem. It’s a 100,000 line problem and that problem doesn’t go away just because the product looks like it works. Just as you’d check a new coder’s work, you have to check AI’s work. That’s never going to change.”

That shift, from builder to supervisor of machines, is uncomfortable. It cuts at the identity of the engineer. But he is not nostalgic about it. The same instinct that led him to build small tools for his family is now applied at a larger scale. 

If you ask what a young developer should do now, his answer is blunt.

“I would start building a ton of things. Use AI to build everything you can imagine. I don’t care if it’s a game or a SaaS product or a text editor. Learn frameworks and how to read code and while the AI spits it out, edit it in real time. Treat the AI like a coding partner, not as a crutch.”

And beyond code, something else matters more than it used to.

“I’d focus on product, UX, and distribution,” he says. “Now anyone can build. The hard part is building something people actually want.”

Writing code is no longer the whole job. Deciding what to build, and making it useful, matters more.

Balabanskyy is dealing with this shift as it happens. His agency is changing how it works. Clients expect more, tools move faster, and the old model is slipping. He is adjusting while it’s still unclear where things land.

The kid who tried to recreate his grandmother with if-then statements would get it. More tools, less waiting, same instinct.

“I always loved development. I always wanted to create stuff. Now I just get to do it much faster,” he says.

His love of creating stayed the same, he says. It’s just that everything around that joy is changing. And that’s alright with him.