"we have tests"

spoiler alert: any test that was written by an LLM as part of the same task that created the feature is less than worthless.

especially if no human reads AND fully understands the tests.