About

Moon Hyeonggeun. Twelve years in QA.

I don’t write the code. It does, when I ask. What I do is find where the thing breaks, and that instinct works the same whether a person wrote it or a machine did.

It works better now, if anything. Machines get it wrong much faster, and much more plausibly.

These days

I set it going at night and go to bed. In the morning sixteen of them have finished.

I used to do all of this myself, by hand. It was hard on the wrist.

What’s here

Four things.

• Built — eight tools I made and run. No demos. If it doesn’t run, I write down that it doesn’t.
• Teardowns — repos, articles and videos taken apart and judged. About one in three ends in “I didn’t take anything.”
• Writing — the notes that outgrew a commit message.
• Just out — a stockpile the machine fills daily and takes apart once. It is the only section I don’t write — what I pick out of it becomes a teardown.

What isn’t here

Work. The QA automation I’ve built for a living is several times this list, but the subject belongs to my employer, not to me. Where I’ve withheld something, I say so.

A list of wins. Headroom only ended after its conclusion had reversed twice, and in Prime Agent the best thing I took was something I had already written down five days earlier and forgotten. That’s what the work actually looks like.

Why write it down

Do QA long enough and you learn that a judgment you don’t record is a judgment you will make again. That happened to me badly enough once that I pulled every link out of two years of session logs and made the result searchable. The teardowns here come off that list.

That list is mine alone to read. This site is the outside of it — anything not left in a searchable form ends up known only to me.

Getting in touch

Repos are behind the GitHub link, bottom left. If you’re actually going to use any of this, an issue is the fastest way to ask.