My Journey with AI

The progression, the WTF moments, the crises, and how my thinking evolved.

How It Started

I am a non-technical person. My biggest coding achievements before last year were little programs I wrote in Pascal in high school and some basic dBase queries I learned at McKinsey. I might have even written some Excel VBA, but I hardly remember.

And here I am in February 2026 — I've "written" over 420,000 lines of code across 8 production projects serving real users. To put that in perspective: in the "old times," that would have been years of work for a team of developers.

So how did I end up here? First there was inspiration. I was passively following the AI revolution, mostly reading news and interacting with ChatGPT or Claude. Then a friend of mine, David, who runs a VC, told me he started "vibe-coding" applications. He didn't just tell me — he showed me two apps he had vibecoded.

That made me think: I'm a manager and operator building an AI-native company, so I need deep understanding of the technical side. And this knowledge is now available. I asked our one-man-band CTO/CPO Justin to help me start... and the rest is history. My journey began.

Critical Milestones & WTF Moments

1. Content Overload → Building

I started by watching YouTube videos. A lot. On one hand I was super excited — most were very cool. But I quickly realized two things: there was already too much content, and I didn't know which was good or relevant for me. Second, even when I watched valuable content, I got pumped but didn't know EXACTLY what to do next. Someone showed me what they did, it was great, but it wasn't answering my precise needs.

The answer: I built an efficient system for filtering content. As opposed to only consuming, I started talking to AI. And most importantly, I cut watching and moved on to BUILDING.

2. From Wrappers to Deep Work

When I started building, I quickly zoomed through the wrapper layer. There are amazing apps like Lovable that make the first step super easy — you write a prompt and magic happens. But I quickly wasn't satisfied. After the initial WOW moment, the next stages weren't so exciting. I couldn't get apps working the way I wanted, had problems solving bugs. It felt like building with Duplo blocks (which get you a house) as opposed to classic Lego bricks (which give you much closer to what you want).

I also realized these apps lock you to their own infrastructure and limit your learning — you end up just prompting without understanding what's happening below. So I made the jump and immersed myself in Claude Code (shoutout to Justin, my Yoda master).

3. Commitment to One Ecosystem

Working with Claude Code (starting in July 2025) has been a constant wave of WOW moments, with some moments where I felt stuck and like I wouldn't be able to get through this. But committing to this one ecosystem (which seems to be the leading one for coding and practical solutions anyway) allowed me to focus. I stopped consuming other content, stopped switching models. That gave me structure — each project was compounding knowledge.

The Progression

I track my AI capability on a 10-level framework. Here's how my journey unfolded:

May 2025 — Level 2: AI-Assisted Worker

Started using Claude, first experiments, learning the basics.

August 2025 — Level 4: Workflow Integrator

First real integrations working. Started building production systems.

October 2025 — Level 6: AI System Designer

Exceptional velocity — API-first mindset, multi-agent design. 3 months from Level 2 to 6.

December 2025 — Level 6 (High-End Mastery)

Depth over breadth — built Cortex platform with meta-agents, 90+ endpoints, 50+ tools. Architecture score: 9/10.

February 2026 — Level 7: AI Product Builder

8 projects, 420K+ LOC, 7 domains, production systems serving real users. Composite score: 8.0/10. Strongest dimension: AI Meta-cognition (9/10).

The jump from Level 2 to Level 7 took 9 months. The key: intense focus, compound learning, and commitment to one ecosystem. No traditional development background required.

What I Love About AI

  • Learning superpowers — It tells you exactly what you need to do, gives you all the resources.
  • Hyper-personalized — Your specific needs, your specific context, your specific learning style.
  • Massive opportunities — Things that were impossible for non-technical people are now possible.

What Worries Me

  • Self-reinforcement at scale — The smartest people are discovering patterns, sharing results. AI is watching, analyzing, immediately implementing. I started feeling like I'm in a knowledge race I can't win even in the short/mid-term.
  • Something will go wrong — There's a high probability something will go terribly wrong. It's out there in the public now.
  • Margin compression — The marginal value of knowledge work is being driven to zero at a rapid pace. Analyst jobs are gone. The compression is moving up the ladder — experience in various areas eaten by models.
  • No place to hide — The agentic system means even niche spaces can be eaten up. It's not like Walmart or Amazon changing retail where some niches survived with high margins — moving there wasn't worth it for them. For AI it's different: marginal cost of taking over any niche is close to zero.
  • How to prepare my children — I genuinely don't know.

But frankly — I don't control those things that worry me. All I can do is enjoy the ride while it lasts and, worst case, delay my "professional obsolescence" for as long as possible.

If I could do it — so can you.

I'm here to help make your journey easier by sharing my mistakes, my learnings, and putting it all into structure so you don't have to ask yourself "where do I begin?" or "what's next?"

Ready to jump? Fasten your seatbelts and enjoy the ride!