I'm in My Fifties. I Started Coding AI Projects. Here's What I Learned.

I'm in my fifties. My job is relationships. I've spent 20 years building trust with Japanese enterprises.

Most business leaders my age read about AI. We attend the conferences. We nod at the strategy decks. We worry about becoming obsolete.

I decided to stop nodding. I decided to build.

Why I stopped coding

I should be clear: I'm not a good coder. Never was.

I understand software architecture well. Databases, backends, frontends, APIs. I've built all of those over the decades. But I was slow. My experience was PHP, MySQL, JavaScript. I didn't have the time or energy to learn modern stacks, and I didn't have the speed to execute efficiently with my existing skills.

The gap between what I could imagine and what I could build had become too wide. So I stopped.

Why I started again

Like many people, I started playing with ChatGPT and felt its potential. But I saw more than a clever chatbot. I saw a coding partner.

Before "vibe coding" had a name, before Andrej Karpathy made it a movement in early 2025, I'd already committed my first AI-written code: September 7, 2024. A WordPress plugin, built over a couple of days.

Then I started using Cursor with Claude Sonnet. Early days. A lot of hand-holding. One step forward, two steps back. But even then, it was magical. And the tools improved every week.

AI flipped the equation. I could direct an efficient coder with confidence. My architectural knowledge became leverage instead of limitation.

Building for my family

In November 2024, I built a script to help my daughter automate her photography uploads. Photos in a folder would get tagged and described by AI, then uploaded automatically to her photography site. She earned her first commissions from those photos.

Around the same time, my son Kenzo was studying web development. His curriculum was solid, but they weren't teaching AI tools yet. So we spent a few weeks building a web app together. Having something published and working was eye-opening for both of us.

From then on, it was a whirlwind. New models, new tools, building apps. While the world was talking about AI, I was building things with it.

What I built

The project I'm most proud of: the GAM Forecast Tool.

This isn't a prototype. It's a working application with the logic complexity of an enterprise platform, running with the privacy and speed of a local utility. The numbers: 200+ days from concept to version 4. Nearly 200,000 lines of code. 1,375 commits. A full database structure, web UI, and AI assistant that understands sales forecasting.

It automates work that used to take 8 hours. Now it takes 5 minutes. Sensitive data never leaves my machine.

I built this. With AI as my partner, but I built it.

What changed

When we have Agentforce trainings now, I understand what's underneath the demos. Not because someone explained it. Because I built something myself.

Having hands-on experience building with AI gives me clarity most people don't have. I can position our solutions to customers with genuine confidence. I can explain to my parents what AI actually means for jobs. I have fact-based opinions, not impressions from headlines.

A casual ChatGPT user only glimpses what's possible. I've seen more.

The point

Get your hands dirty. Try things you're not comfortable with. Have fun.

This is not just about companies making more money. It's about agency. It's about refusing to let your experience become a relic.

The tools are here. The barrier is gone. You don't need to become a developer. But you must stop being a spectator.

You don't need to build to have AI opinions. But building changed mine.