Debugging in the Dark: How Clinical Thinking from Interventional Radiology and KIRO AI Helped Me Fix a HarmonyOS Bug

A few weeks ago, I ran into a frustrating problem.

I was working on a mobile application used by large-scale industries, and we started receiving multiple complaints from users on HarmonyOS devices. The issue was consistent — something was breaking — but I had one major limitation:

I didn’t have a physical HarmonyOS device to reproduce the issue.

That meant I was debugging in the dark.

An Unexpected Conversation

Around the same time, my wife had just been transferred to a new hospital unit: Interventional Radiology.

One evening, while I was picking her up from work, I asked what her new unit actually does.

She explained it in a simple way:

“Instead of performing open surgery, doctors use real-time imaging like CT scans or ultrasound to guide instruments inside the body.
They can see exactly where they are going — and fix the problem with precision.”

That stuck with me.

The Moment It Clicked

As I was driving home, my mind kept jumping between two things:

• My debugging problem (no device, no reproduction)
• Her explanation of scan-guided precision

Then it hit me.

“What if I could scan my codebase the same way they scan the human body?”

Instead of blindly guessing where the issue was, I needed a guided, diagnostic approach.

Applying the Idea

So I tried something different.

I used kiro-cli alongside Amazon Q to deeply analyze the project — not just at the surface level, but across dependencies, platform-specific conditions, and compatibility layers.

kiro-cli helped me scan the structure of the project and uncover hidden inconsistencies, while Amazon Q acted like an intelligent assistant — helping me reason through platform-specific behavior, identify suspicious patterns, and validate assumptions quickly.

Just like a radiologist uses imaging to pinpoint a problem area, I used these tools to:

• Identify HarmonyOS-specific incompatibilities
• Trace execution paths that behave differently on that platform
• Detect subtle issues that wouldn’t show up in standard testing
• Validate fixes faster with AI-assisted reasoning

And there it was — the root cause.

A small but critical platform-specific issue that I would’ve struggled to find manually.

The Result

I fixed the issue without ever touching a HarmonyOS device.

And more importantly, I changed how I approach debugging:

• Less guessing
• More guided analysis
• Leverage AI tools effectively
• Think like a “diagnostic system,” not just a developer

What I Learned

In Interventional Radiology, precision comes from seeing clearly before acting.

The same applies to software engineering.

Modern tools like kiro-cli and Amazon Q are not just productivity boosters — they are like diagnostic scanners for your codebase.

Sometimes, the best solutions don’t come from more effort —
they come from seeing the problem differently.

Final Thought

Inspiration doesn’t always come from within your field.

Sometimes, it comes from a conversation in a car… after a hospital shift.