That’s right, you heard right! At long last, we can stop talking about this _ridiculous_ practice. We can stop chastising ourselves and pretending that you can “test” something that doesn’t exist yet.

Test-Driven Development didn’t fail to catch on because it was a bad way to write code; it failed because its name doomed it to be misunderstood from the start.

In this talk, we’ll drop the “testing” mental model, and reframe this practice as _Continuous Specification_, a practice that actually makes sense of what TDD was trying to achieve for all these years. It also solves an existential problem for coding with AI: how to convey unambiguous expectations for our code, so we can move fast _without_ turning our codebase into a haunted house.

(If it isn't clear from the abstract, I am absolutely a TDD advocate. This is a tongue-in-cheek way to reframe the conversation and show how essential this practice is as we move into coding with agents more than in IDEs.)