From Flashcards to AI #3 | Building the Cards Was Harder Than Building the Application

When the learning engine was finally ready, I felt a sense of accomplishment.

• The application could display long questions.
• It could handle detailed explanations.
• It could manage multiple decks.
• The interface worked.
• The deployment worked.
• The study workflow worked.
• Everything seemed ready.

Except for one small detail.

• I had nothing to study.

At first, this sounded almost ridiculous.

After all, I was preparing for a Microsoft certification.
Surely there would be plenty of learning material available.

And there was.

• Microsoft Learn.
• Workshop materials.
• Practice assessments.
• Documentation.
• Notes.
• Videos.
• Examples.
• Labs.

The problem was not the lack of information.
The problem was the opposite.
There was too much information.

Very quickly, I realized that simply consuming content would not be enough.
Reading a module and moving on felt productive. Sometimes it even felt like progress.

But a few days later I would discover that I remembered much less than I thought.
I needed a different approach.
Something active.
Something that forced me to interact with the material.
Something that would make gaps in my understanding visible.

That was the moment when I started creating my own learning cards.

At first, the process felt painfully slow.

For every topic, I had to ask myself:

• What is the key concept?
• What question could test this knowledge?
• What answer would I expect?
• Why is that answer correct?
• What common misunderstanding should I remember?
• What additional context will I need three weeks from now when I review this again?

Creating a single good card often took significantly longer than reviewing it later.

Without realizing it, I was no longer just collecting notes.

I was building a personal knowledge base.

One card at a time.
One topic at a time.
One mistake at a time.

The source material came from many places.

• Microsoft Learn modules.
• Official practice assessments.
• Workshop exercises.
• Azure documentation.
• Personal notes.
• Screenshots.
• Experiments.

Every useful piece of information eventually found its way into the system.

But something interesting happened.

The process of creating the cards became part of the learning itself.

• Sometimes I discovered that I couldn’t explain a concept clearly.
  That usually meant I didn’t understand it well enough yet.
• Sometimes I realized that two Azure services looked very similar.
  That became a new card.
• Sometimes I found myself writing explanations that were longer than the original question.
  That became another card.

As the number of cards grew, something else changed. The application stopped feeling like a software project. And started feeling like a study companion.

Every evening I would add new cards.
Every morning I would review old ones.
The feedback loop became incredibly effective.

Learn. Convert. Review. Refine. Repeat.

Eventually, my Excel workbook contained hundreds of entries.

Questions. Answers. Explanations. References. Images. Examples.

At that point, the cards had become a deck.
And the deck had become a serious learning resource.

Fortunately, I already had another tool waiting for me.

A converter I had previously written could transform the structured Excel data into the JSON format required by the application.

What had started as rows in a spreadsheet could now become interactive learning content.

A few clicks later:

• the JSON files were generated
• the images were prepared
• the content was uploaded
• the decks were deployed

And suddenly, the system was alive.

Not as a project.

Not as an experiment.

But as the primary tool I would use to prepare for one of the most challenging certifications I had attempted so far.

I thought the difficult part was behind me.

I was wrong.

A much bigger challenge was about to arrive.

And it came from a place I wasn’t expecting.

A routine team meeting.