Building the Future: An Interview with ADI's Founder
When most developers were still trying to optimize AI assistants that help write code, Ihor Herasymovych was asking a different question: what if AI didn't just assist developers, but autonomously handled entire product workflows?
We sat down with Ihor, ADI's founder and a 13-year IT veteran with two CTO stints and R&D leadership experience, to discuss the origins of ADI, his contrarian views on AI development, and why he believes the industry has been solving the wrong problem.
The Origin Story
Let's start at the beginning. What's your background, and what led you to build ADI?
I've been in IT for over 13 years, served as CTO twice, and led R&D in ventures where we had to build things fast. I've been an early AI adopter, experimenting with almost every AI coding tool that came out.
The "aha moment" came when I was using Claude Code. I realized that for a lot of tasks, I wasn't even looking at the responses until the finale. I'd give it a task, walk away, and come back to completed work. That's when it clicked: if I'm not actively participating in the development process, why am I even in the loop?
When did you start building ADI?
May 30th, 2025. It started as a side project—something I was exploring nights and weekends while working on other things.
What changed in October that made it full-time?
Three things converged: interest from people in the industry, conversations with investors and ventures, and the results from our private beta. The cost and time reduction on tasks was huge. We saw real validation that this approach works.
Why GitLab First?
Most AI dev tools target GitHub first. Why did you choose GitLab?
Personal preference and knowledge. I understand how their infrastructure works, and the self-hosted option was important from day one. GitLab's architecture aligned better with what we were building.
Is GitHub support coming?
Absolutely. It's in our 6-month roadmap. GitHub support, additional task sources, and expanding platform compatibility are all on the list.
The Philosophy: Replacing, Not Assisting
Your approach seems fundamentally different from other AI coding tools. What's your philosophy?
Here's where I diverge from most of the industry: AI should replace the developers who just move buttons or do non-mind-consuming things.
Everyone else is trying to solve these enormous problems or achieve ideal results or assist the developer. We're not doing any of that. We're not assisting developers—we're assisting the product by having a 100% autonomous flow.
That's a bold stance. Can you explain what "100% autonomous flow" means in practice?
It's actually simple: you create a task in any task source, and it will automatically be completed if AI decides that task is AI-fixable. No developer intervention needed during the process.
When does a human need to step in?
Reviewing code, approving deployments, and business decisions. Those are the touch points. But the actual development work—reading requirements, writing code, running tests, iterating on failures—that's fully autonomous.
BYOK and Self-Hosted: Both Worlds
ADI uses a Bring Your Own Key model. Why not just host everything?
We actually offer both options, and each serves a different need.
BYOK gives users complete cost transparency and control. You see exactly what you're spending on AI API calls. You own your keys, your data never touches our infrastructure in ways you don't control.
What about the hosted option?
Self-hosted will be available in the near future. It's for companies that care deeply about privacy—companies that don't want to leak even their task names to external services. We don't plan to access that data anyway, but self-hosted adds an additional layer of certainty.
It's also for organizations that don't want to cross certain infrastructure boundaries or have specific compliance requirements.
The Technical Journey
You mentioned trying almost all AI coding tools before building ADI. What didn't work about them?
They're all trying to solve too big of a problem. They want to be the perfect assistant, the ideal pair programmer. That's not what the market needs.
What I saw was: I don't need an assistant. I need autonomous execution. The tooling didn't exist, so I built it.
What were the biggest technical challenges in those first months?
Honestly? Not the AI orchestration or GitLab integration—those were challenging but straightforward. The hardest part was moving from a working side project to a usable platform. Taking all the pieces that worked in isolation and merging them into something coherent that others could use.
Current State and What's Next
Where is ADI today?
We're in public beta. All functionality works, but we're calling it beta because we're 100% validating product-market fit. We want to ensure we're solving the right problems for the right users before we declare it production-ready.
What's your 6-month vision?
GitHub support is priority one. Then additional task sources—we want to integrate with more project management tools.
Self-hosted deployments need to be rock solid. Community building is huge—we need users sharing their experiences, contributing ideas, helping shape the product.
And then the business side: starting a funding round or establishing permanent revenue flow to fuel growth. PR and marketing to get the word out.
What do you want ADI to be known for?
The platform that proved autonomous AI development works at scale. Not as a concept, not as a demo, but as a real tool that ships real features to real production environments.
We're not here to make developers 10% more productive. We're here to eliminate entire categories of work that don't require human creativity or judgment.
Final Thoughts
What would you say to developers who are skeptical about autonomous AI development?
Start small. Give it a task you'd normally spend an hour on. Walk away. Come back and review the result.
You'll realize two things: first, the AI probably did it competently. Second, you just got an hour of your life back to work on something that actually requires your unique human skills.
That's the future we're building. Not replacing developers—replacing the parts of development that waste developer potential.
ADI is currently in public beta. If you're interested in trying autonomous AI development, visit the-ihor.com or check out our documentation.