Catch it early,
or pay for it later
Good results come from the right questions — asked as early as the spec, not as late as the release.
Process scales with the stakes
A spec exists so the test suite has something to check against besides itself. Without one, an agent will happily generate tests for whatever it decided the feature does — plausible, green, and validating the wrong thing.
If I do it twice, I script it — and increasingly, an agent does
Automate the repeatable, save judgment for what actually needs it. Three years of QA taught me this the hard way — and today that means building agentically by default, not bolting AI on as an afterthought.
Zero trust by default
I don’t grant trust automatically — not to a request, not to a green check. A passing suite can still be a false positive, so I lean on a standardized QA process, not just the run’s result, before I call something verified.
Curious enough to break it
I’m building toward DevSecOps and application security on top of a habit QA already gave me: chase the question nobody thought to ask yet.
Shipping isn’t the finish line.
I care what happens after: does it hold up, does someone actually want to use it, does the person maintaining it thank me or curse me.
Open to QA, automation and security roles
Fastest reply by email — happy to talk through the work in a short call too.