Green CI Tests Confirm Correctness, Not Improvement — Here's the Difference

A widely shared developer essay argues that passing test suites only confirm code correctness, not whether a change actually improves the system. The green bar in CI verifies that inputs map to expected outputs, but it does not measure outcomes like latency, user activation, or agent behavior. The author distinguishes Spec-Driven Development, which ends at a passing build, from Hypothesis-Driven Development, which starts with a predicted outcome and ends with measured validation. Frameworks from Thoughtworks and PMI have long addressed outcome measurement, yet most engineering teams treat test conformance as a proxy for improvement. The core argument is that correctness is necessary but not sufficient for progress, and the two questions — did it work as written, and did it make things better — require separate, deliberate instrumentation.