Testing Practice: Tests That Can Fail by Samuel AderemiTesting Practice: Tests That Can Fail by Samuel Aderemi

Testing Practice: Tests That Can Fail

Portfolio project
Samuel Aderemi
Across a multi-surface platform with around 1,900 automated tests, I don't trust a test until I've seen it fail.
On one feature, all 94 tests passed while the feature did nothing at all: the check was reading its own log entry. Since then, I remove the thing a test guards, watch the test fail, and put it back. The client singled this out as the practice he valued most.
I check claims against the live system, not from memory: schema read back before writing to it, byte-level hashes compared between source and what the server actually serves, and page measurements taken in a real browser with Playwright.
Every deploy is kept as a rollback, and no secrets go in page code, logs or version control. I once stopped a deploy after proving the repository and the live build had drifted apart by hundreds of lines. The client's words: "That call saved live work."

0 likes · 0 views · Nov 17, 2026

~1,900 tests, each one proven able to fail. Live-system checks, rollback on every deploy, and a stopped deploy that saved live work.

Tags