Writing you can check.
Every number here is traced to a named publication. Where the evidence does not exist, we say so rather than reaching for a statistic that sounds right.
RSS feed · 10 articles
A passing run is not a verified run
“Pass” is doing two jobs on most dashboards: the step completed, and the step confirmed something. What the standards say a result is, where unchecked steps come from, and the one number to ask for before a sign-off.
EngineeringWhen you need a real browser, and when headless is enough
Headless Chromium has been the real Chrome engine since version 112. What a real browser on a real OS still shows that headless cannot, who actually runs which browser, and a decision rule that keeps cross-browser affordable.
EnterpriseTest evidence for regulated releases
A test run has two readers. One wants to know what broke; the other wants to know the record is true. The regulation that defines evidence most precisely, applied to AI-driven execution.
EngineeringHow to evaluate self-healing before you trust it
The failure mode a green result hides, five properties of healing you can trust, and a three-change test you can run against any tool in an afternoon — including ours.
GuidesWhat “plain English” actually means in test automation
The phrase covers four different capabilities, and a tool can honestly claim it while delivering only one. Three questions tell you which you are being sold.
EconomicsWhat test maintenance actually costs
Maintenance is three different activities budgeted as one. What the research says about each, and a four-number measurement any team can take in a single release.
EngineeringWhy tests break: a look at locator strategy
Over 70% of web test breakages trace to element locators. The published comparison of locator strategies — including how far the best available strategy actually gets you.
EconomicsThe codeless testing market in 2026
Three credible firms sized this market for the same year and came back 60% apart. What causes the spread, what survives it, and which numbers a buyer should use instead.
GuidesRequirements quality and defect prevention
Everyone quotes the 100× figure. The three findings printed next to it are more useful — and they explain why coding agents make specification quality a bigger problem than it was in 2001.
GuidesManual to automated testing: a migration guide
Most automation programs do not fail — they stall somewhere under half coverage while the manual regression bill keeps arriving. What the survey data actually says about that, and a five-step sequence for getting past it.
Pieces are published when the sourcing is done, not on a schedule. If there is something you want covered, tell us.