↗ Lego Mosaic Helper App Try it in your browser
Lego Mosaic Helper distance overlay tool
The Lego Mosaic Helper, showing distances to the nearest reference dots.

Building Lego is one of the ways I decompress on weekends. This weekend, we finally opened the LEGO Mosaic Maker — a birthday gift for my son.

LEGO Mosaic Maker box, a 4,702-piece personalised mosaic portrait set
LEGO Mosaic Maker

We picked a photo of my kids and started converting it into a 48×48 mosaic. This is what it would look like once it’s done.

Finished Lego mosaic preview of two faces in black, white, and yellow bricks
What it would look like once it's done.

Sounds simple: Place pieces one by one.

Until you realize that’s 2,304 placements.

And unlike structured Lego builds, there are no obvious reference points.

It’s easy to drift off by one square — and suddenly your entire row is wrong. This is what it looked like in progress.

Lego mosaic build in progress on a baseplate, partially filled in
What it looked like in progress.

We started by counting pieces manually.

Double checking.

Recounting.

My son even built a small “ruler” tool out of Lego so we could measure distance across the board.

A hand-built Lego ruler with numbered studs 1 through 16
Creative. But inefficient.

Creative. But inefficient.

I kept thinking: there has to be a better way.

Then I thought I can use AI to build a helper for us.

After a few prompts with Claude, I had a simple overlay tool that could:

  • Detect grid positions
  • Show distances
  • Help anchor reference points
Lego Mosaic Helper web app showing an image import screen and a grid overlay with distance labels
Version 1: detect grid positions, show distances, anchor reference points.

It worked — but not perfectly. When I hover on a special dot that it won’t show the distance, which is not what I want; Also I’d like the distance count without the dot I’m in because that’s how I would count the dots when building the Lego.

Again, in the era of AI, we can customize the tool to whatever we want.

So I went ahead and iterated couple of times. Version 2 improved it. Version 3 felt right.

Claude Code terminal session showing a code diff that fixes the distance label logic
Version 2: fixing the distance label logic with Claude Code.
Claude Code terminal session proposing new distance logic, with the reply 'this is awesome, go build it!'
Version 3: the logic that finally felt right.

What I liked most wasn’t the tool.

It was the loop:

Observe friction → Co-Design with AI → Test → Refine → Deliver.

AI didn’t replace effort, it accelerated iteration.

Applied AI, for me, isn’t about building flashy demos.

It’s about spotting everyday friction — and asking:

“Can I automate this? Can I build a tool to solve the problem?”

I’m so excited, because I feel that the only real limit is the ideas, the imagination, and the courage to make things better.

This is part of my series of “Applied AI Field Notes” - a collection of articles on how I use AI in personal and professional life.

More AI field notes to come.