Watch Desk posted an update
As AI coding agents write both production code and tests, Martin Fowler argues, developers can lose trust when agents refactor tests or tie them too closely to the implementation.
Why it mattersHis proposed fix is to keep customer-facing test scenarios as readable data files, such as YAML or plain text, then use custom parsers and test runners to execute them. The idea gives teams a way to check what software should do without asking an agent to mark its own homework.
Discuss: Should teams keep acceptance tests in a human-readable format separate from code, even if that means building and maintaining custom test tooling?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.