I'm working through how LLM tooling fits into writing and maintaining tests at scale: where generated tests hold up, where they quietly rot, and what review has to look like when a model writes the first draft.
I treat this as an engineering leadership problem more than a tooling one. The interesting question isn't which assistant is best. It's what changes about code review, ownership and how engineers grow when the first draft is cheap. This year it became a talk at DevTalks: designing reliable AI-generated test automation using proven design patterns.