The team achieved 100% test coverage.
They celebrated their thoroughness.
The senior developer asked, “What percentage of your understanding is covered?”
And then, “If every line is tested, what remains untested?”
Understanding the Koan
We treat code coverage as a destination, a number that validates our diligence. When that metric reaches 100%, we feel we’ve done our due diligence. But this koan points to a deeper question: what does it mean to truly test software? Coverage measures execution, not comprehension. It tells us which lines ran during our tests, but it cannot measure whether we understood what those lines should do, or what they should never do.
The senior developer’s first question cuts to the heart of this confusion. We can test every line of code without understanding the problem we’re solving. We can achieve perfect coverage while completely missing edge cases, race conditions, or subtle business rule violations. The tests themselves might be testing the wrong things entirely, validating implementation details rather than behavior, or worse, simply confirming that the code does what it currently does rather than what it should do.
The second question reveals an even deeper layer. Even if every line executes during testing, what about the interactions between those lines? What about the states that only emerge after thousands of operations? What about the assumptions embedded in our code that we never thought to question? What about the requirements we never received, the user behaviors we never imagined, the infrastructure failures we never simulated?
This koan invites us to sit with the discomfort of knowing that no metric can substitute for understanding. Code coverage is useful as a floor, not a ceiling. It can show us what we definitely haven’t tested, but it cannot tell us what we have truly verified. The practice here is not to abandon coverage metrics, but to see them clearly for what they are: a measure of execution, not a measure of wisdom.
The real work of testing begins when we ask not “Did this line run?” but “Do I understand what this code is supposed to do, and have I verified that understanding?” That question has no percentage attached to it.