The original headline is: "Tests Pass, Design Breaks: Why TDD Can't Hold the Line on Design Intent"

Short summary

TDD verifies that implementation matches tests, not that tests correctly express design intent. When AI writes code and tests quickly, humans must review design alignment, but lack capacity. The safe_post.py example shows how tests can pass while callers use outdated signatures. Solution: separate machine-verifiable concerns (types, signatures) from human-verifiable ones (design intent alignment).

• •TDD's structure doesn't verify if tests correctly express design intent—only if implementation matches tests
• •AI-written tests are often reverse-engineered from implementation, missing original design goals
• •Separate machine-verifiable (types, signatures) from human-verifiable (design alignment) tasks

Generated with AI, which can make mistakes.