When the World Sleeps, Some Ideas Wake Up

People often ask me what I do when I can’t sleep.
I used to worry about it. Now, I build.

This project didn’t start as a product idea. It didn’t begin with a plan or a business goal. It started with something very simple and very personal—my son.

I wanted him to practice abacus every day. Nothing intense. Just 20 questions. About three minutes. But as every parent knows, building a habit is harder than solving the problem itself.

So one night, without thinking too much about the future, I decided to build a small app for him. Just a quick experiment. I genuinely believed I would be done in a day.

That didn’t happen.

One feature led to another. One night turned into many. And without realizing it, I was no longer just fixing a small problem for my child—I was shaping a real product.

Over time, that simple app evolved into a brain training tool designed not only for kids, but also for adults and for people who want to keep their memory and cognitive skills sharp. Right now, it supports practice with numbers using addition, subtraction, and decimals, with quick daily sessions that don’t feel overwhelming.

Under the surface, it also includes behavioral and cognitive feedback powered by on-device machine learning using TensorFlow Lite. I built both classification and regression models so the app can adapt based on how a user is actually performing, not just how often they practice.

One decision I was very clear about from the beginning was data privacy.

As a parent, I never wanted my child’s data to live on unknown servers. And as a builder, I knew this concern would extend far beyond just kids. So I designed everything to run locally on the device. No practice data is uploaded anywhere. The only optional feature is sharing a practice result sheet—so it can be viewed on another device or shared with a parent or teacher. Nothing else leaves the phone.

Later, I added visual abacus animations so serious learners could review exactly where a mistake happened, down to the borrow and carry steps. That was one of the most satisfying additions, because it turned practice into real understanding.

At some point, I noticed something unexpected—I no longer felt the urge to add more features. For the first time in a long while, the product felt complete. That quiet sense of completion was deeply fulfilling.

Now, as I slowly begin brainstorming my next late-night project in the space of computer vision and machine learning, I wanted to pause and document this journey.

Not as an announcement.
Not as marketing.
Just as a reminder to myself that some of the most meaningful things don’t begin with ambition. They begin with responsibility, curiosity, and a few sleepless nights.