Writing — page 2 of 2
Which technique finds which bug, how to catch a requirement that cannot be tested, and where AI earns its place in a QA workflow.
Generation used to be the bottleneck in QA automation. Review is the bottleneck now, and a rising coverage percentage doesn't show that the shift happened. This post traces where the hours went and which judgment calls a reviewer still can't hand off.
Requirements defects account for roughly half of all software defects, and most of them trace back to five phrases that sound specific but aren't. Each one gets a concrete replacement you can drop into a spec today.
Both techniques test the same input, but they answer different questions: one is about which group a value falls into, the other is about the edge where a group ends. Worked examples of each, and the rule for combining them without doubling your case count.
An import that fails at row 40 usually traces back to one of five predictable mistakes: merged cells, a missing ID column, mismatched steps. This post lays out the flat, single-sheet shape that TestRail, Qase, Zephyr, and Xray all accept without a rewrite.