There are reasons why I measure code quote input unquote via testing and static code analysis:

1. re-establishes mind mapping of the code base 2. separates out noise from signal. 3. makes it a breeze to refactor to reduce code complexity.

Notice anything yet?

If I replace the AI code input with my own curated techniques from known and measured code sources in long-term I save more time than those who rely upon AI vibe coding.