Three hours, per feature.
One feature equals a half-day of a QA engineer's time, just for writing test cases.
The pain of manual TC writing — endless rework, quality gaps, three hours lost per feature — gave way, one spring afternoon, to something quieter: a pipeline that reads a spec and quietly gets to work.
For most of the last year, the writing of game test cases in our studio took three hours a feature, give or take. A senior QA engineer would sit with a Confluence page — sometimes a PDF, if she was unlucky — and type out three hundred test cases by hand. She would then reread the spec, and do it all over again when something changed, which something always did.
Then, one afternoon in the spring of 2026, she handed the spec to a pipeline. A designer read it; an inspector reviewed the design; a writer filed the cases to Google Sheets. A first reviewer flagged problems, a fixer fixed them, a second reviewer went deeper, a second fixer went deeper still. Forty minutes later, three hundred test cases arrived — on par, she said, with what a smart three-year senior would have written. She had not touched her keyboard.
She had not touched her keyboard.
One feature equals a half-day of a QA engineer's time, just for writing test cases.
Manual test cases miss edge cases and forget state transitions.
A single spec revision means re-reading and rewriting dozens of test cases.
Just provide a spec link and a Google Sheets link. The rest is hands-off.
Measured on a real 300-test-case feature run.






"I used to write three hundred test cases by hand. Now I mostly drink coffee."
Phase I · Now.
Phase II · Soon.
Phase III · Future.