// Building a Statistics Tool with AI

Let me start by saying: I'm not an engineer.

I started as a systems engineer at NTT, then worked on games and web services at Square Enix and Drecom. I've built web applications on my own and have touched infrastructure end to end. But I'm fundamentally someone who "has tried to understand what development teams do" — a direction and management person, not a professional coder.

I'd never built interactive data visualizations for the frontend professionally. Honestly, I never thought I could build something like this.

This is the story of how someone like me built a statistics learning tool.

// Why I Built It

I started studying for the Grade 2 Statistical Examination and opened the textbook. Got stuck on confidence intervals. I could read the text, but it wouldn't click. Kept going back, feeling like I understood, then the fog would return.

Then I thought: "If I could just see it move, I'd get it instantly."

My years in game and app development taught me that interactive experiences transform concept understanding. Move a slider, watch the graph change, see the numbers shift — that's what creates the "aha" moment. It reaches people that text and formulas alone can't.

But such tools didn't exist. At best, a few static diagrams with arrows, explaining "if it moved, it would look like this." Barely anything you could actually interact with.

So I decided to build one.

// Couldn't Have Done It Without AI

I didn't have the technical skills to build this. Before, this would have ended at "I want to build it but can't."

But now there's AI.

I worked with Claude (Anthropic), iterating through conversations — not just "prompt in, code out," but continuous dialogue about "I want it to move like this," "is this design decision right," "will this break later?"

As a longtime director, I could articulate requirements and evaluate output reasonably well. The experience of trying to understand technology in development teams carried over directly to AI collaboration.

// "Having AI Build It Is Easy" — Half Wrong

A digression, but this needs saying.

Give AI instructions and code comes out — looks easy. But in practice, it was a constant stream of decisions. Is this design correct? Will this implementation survive future extensions? Is this feature really necessary? What to include, what to cut? You have to make all these calls yourself.

AI sometimes says "sure, I can do that" and runs in the wrong direction. Evaluating and stopping it is the human's job. Without understanding technical structure to some degree, you can't make these judgments.

If you can't evaluate whether "ES modules or not" or "add a dependency library or not" is the right call, you'll just accept whatever AI proposes. The result: something that works but breaks easily, works but nobody can touch in three years.

I managed this because my experience trying to understand technology in development gave me just enough foundation for judgment. Conversely, approaching AI collaboration with zero foundation is harder than it looks.

// Where This Tool Sits

I have absolutely no intention of replacing textbooks. If you want systematic learning, use textbooks and problem sets. You can't practice calculations here, can't prep for exams.

But if you open it at that moment when you think "I only vaguely understand this concept" — that's enough.

// Finally

The motivation was simple: I built it because I wanted it myself.

And while building it, I discovered the fascination of statistics. Trying to visualize a concept forces you to truly understand it — building became learning.

That's why I want to keep learning. Statistics runs deep; there's so much I don't know. I want to keep adding content, making it more intuitive, incorporating more gamification — building something that makes people feel "statistics is actually interesting."

All code is on GitHub, open source. Feel free to modify it. Want to use it in class, embed it in your site, add new concepts — if it helps education, that makes me very happy. For commercial use, please consult per the CC BY-NC 4.0 license.

If it reaches someone who thinks "if only I could see it move, I'd get it" — that's enough.

Statistics is actually really fascinating.