How I spent 12 months building a custom OCR engine just to win the "Homework War" at my kitchen table

Hey everyone,

After my last post, I realized that I focused a lot on the what, but not enough on the how and the why.

If you’re a parent, you know the "Homework War." It’s that 7:00 PM struggle where a boring worksheet becomes the villain of the evening. I’m a dev, but I’m also a dad, and watching my daughter struggle with spelling lists was the spark for WordyKid.

I thought building this would be simple: take a photo, run it through a standard OCR API, and plug it into a game engine. I was dead wrong.

The "Kid Handwriting" Challenge
Standard OCR is great for clean, printed business documents. It’s terrible at:

Messy, half-erased pencil marks.

Blurry photos taken by tired parents under kitchen lighting.

Non-standard worksheet layouts where text is scattered everywhere.

We spent over a year building a custom architecture to handle this. We had to create a bridge between computer vision and a dynamic game engine that doesn't just "read" the text, but understands the structure of a learning assignment.

The Result?
The "villain" (the boring worksheet) is gone. Now, my daughter just snaps a photo, and 30 seconds later she’s playing a game that actually teaches her the exact material she needs for class.

I’d love to hear from other founders:

How do you balance "boring" technical depth (like custom OCR) with "flashy" marketing messaging?

If you’re a parent, what’s the #1 thing that makes you close a "parenting tool" site immediately?

Check out what we built here:
https://wordykid.com/

Would love to chat in the comments!