Quality is the degree to which software meets its requirements without defects. It is not just a final check on code: unclear requirements, unrealistic test data, unsafe deployments, and late monitoring signals can all let defects travel downstream. section 1.2.5 section 1.2.2
Build quality into each phase by asking:
- Intent: Are requirements clear, testable, and complete? section 1.2.5
- Build and test: Do tests prevent defects, and are failures investigated rather than worked around? section 1.2
- Deploy and observe: Are releases safe and reversible, and do monitoring signals reach the right people in time? section 1.2.5
A defect caught too late
A new API endpoint reaches review before an event-sourcing requirement comes to light. A performance defect then appears in production. The first was a defect in intent; the second was a defect in design. Both prompt the same question: what would have exposed the problem earlier? section 1.2.2
Quality belongs to everyone who changes and ships the system, not only reviewers or QA. Catching a defect during implementation avoids the handoff and context switching needed to fix it later. Pair programming, test-driven development, and automated linting are ways to catch problems early; a developer who merges AI-generated code also owns its edge cases and integration assumptions. section 13.4.3
When defects recur, locate where they enter the delivery loop and where they are first detected. Strengthen the earlier signal rather than relying on a later phase to catch the same problem again. section 1.2.2 section 1.2.5









