>unit-tested

Some of this is less to do with Claude vocabulary and more to do with the expectation that Claude justifies it's work. That expectation (probably) came from reinforcement learning.