TL;DR: I shouldn't be writing code. As a CEO, my job is strategy. But 12 months ago, I realized I was managing a "Black Box" I didn't understand. So, despite having no background in engineering, I forced myself to build 100 AI products—including a social network for cats—using nothing but AI agents. The Lesson: If you are a manager who "uses AI" but has no "Hand-Feel" (Teigokoro) for the code, you are already obsolete.
James here, CEO of Mercury Technology Solutions.
Tokyo - February 16, 2026
By all traditional logic, I should have stopped "working" years ago.
I run a tech company. I have a team of brilliant engineers. My job is to point at the mountain; their job is to climb it.
But for the past 8 months, I haven't just been pointing.
I have been building.
I have personally built 100 AI products.
I’m not talking about writing specs for my team. I mean opening the IDE (Cursor/VS Code), prompting the AI, and deploying the code myself.
I built "CatsBook" (a social network exclusively for felines).
I built a real-time interactive webinar platform.
I built a Tower Defense game.
Why? Why would a CEO exhaust himself doing the work of a junior developer?
Because I saw a crisis. And that crisis is likely happening in your company right now.
1. The Crisis of "Hand-Feel" (Teigokoro)
Early last year, ChatGPT was everywhere.
I asked my engineering team: "Are we using AI?"
They nodded enthusiastically. "Of course, James. We use it to write code every day."
I felt good. We were innovative.
But then I looked at the metrics.
If we were using AI, why hadn't our delivery time halted? Why hadn't our output doubled?
The results didn't match the hype.
I realized I couldn't manage this problem because I lacked "Hand-Feel" (what the Japanese call Teigokoro / 手頃).
I didn't know how AI coded, so I couldn't tell if my team was using it effectively or just using it as a fancy spell-checker.
I was flying a plane with no instruments.
So, I decided to jump.
I committed to 7 days of "AI Hell." 14 hours a day. Just me and the Agent.
2. The Rubik's Cube Epiphany
I decided to start hard: Build a 3D Rubik's Cube solver.
I have zero background in Matrix Algebra or 3D rendering.
I failed for 6 days straight.
The AI would generate code. I would run it. The cube would break.
Red squares appeared on black faces. The logic was impossible.
To a non-coder, the scripts looked like alien hieroglyphs.
On Day 7, I had an epiphany:
"I don't need to know the syntax. I need to know the LOGIC."
I stopped trying to debug the code. I started debugging the Thinking.
I treated the AI not as a tool, but as 50 Expert Consultants sitting in a room.
I asked:
- "Explain this function to me like I am a 5-year-old."
- "Walk me through your calculation order, step-by-step."
It turned out, the math was right, but the logic of the cube's orientation was backward.
I literally printed out the flat pattern on paper, folded it, and physically proved the AI wrong.
Once the logic was fixed, the code fixed itself.
3. Action > Study (The Inversion of Learning)
Most people learn linearly:
Step 1: Learn Python Basics $\rightarrow$ Step 2: Learn Databases $\rightarrow$ Step 3: Build App.
This is too slow for 2026.
My method was inverted:
- Project 1 (A Choice Game): I learned how to use an IDE and deploy to the cloud.
- Project 2 (Image App): I learned how to connect APIs.
- Project 3 (Leaderboard): I learned SQL databases.
I didn't learn to build. I learned by building.
If you wait until you are "ready," you are dead.
4. Don't Let Inertia Kill You
Inertia is great 99% of the time. It lets us brush our teeth without thinking.
But in the 1% of times when the environment shifts violently (like right now), Inertia is a suicide pact.
Amazon survives because they turned Innovation into their Inertia.
I tell myself every morning:
"If I do the same thing today as I did yesterday, and expect a different result, I am insane."
That is why I forced myself to build 100 products. I had to retrain my brain to make AI creation as instinctive as breathing.
Conclusion: You Have 50 Experts. Use Them.
This isn't a "Learn to Code" story. This is a "Commercial Survival" guide.
If a non-technical manager like me can build a Facebook-level clone in a few days, what is your excuse?
My advice is simple:
Imagine you have 50 PhDs sitting next to you.
- Don't know the financial model? Ask the AI to teach you ROI.
- Don't know the code? Ask the AI to visualize the logic.
Do not ask "What if it hallucinates?"
Ask it again. Ask it differently. Ask until you understand.
Action is always greater than the perfect plan.
Go build something. Today.
Mercury Technology Solutions: Accelerate Digitality.
