Sunday 2026-06-28. intent-mail repo, fresh Stryker baseline run. The coverage gate reported green: 69.09% line coverage. Three seconds later, mutation testing reported 24.88%. The rules engine—the code that actually mutates user email—reported 0.00%.
That zero is the story.
The Gap
Coverage counts lines. Mutation testing counts assertions. When a 535-line end-to-end suite exercises a piece of code but doesn't assert on its internal logic, coverage sees a line executed and calls it a win. Mutation testing inverts a single boolean operator in that line, runs the suite again, and if the outcome is the same, marks that mutant as survived. The engine had 301 mutants and zero of them were killed—not because there were no tests, but because the tests that ran the code never asserted on the code's decision logic.
Here's the real shape of it:
