glossary · end-to-end testing

Vibe testing

what is vibe testing?

Vibe testing is the practice of verifying software that was generated rather than written line by line, by checking that user journeys still work instead of reviewing the code that changed. It is the counterpart to vibe coding: when changes arrive faster than anyone reads them, behaviour becomes the only thing left worth checking.

The problem is specific to the way AI-built applications fail. A prompt-driven change can rewrite a component nobody opened, and everything looks fine because the page still renders. What broke is three screens away, in the flow that used to lead to a purchase.

Code review does not scale to this and neither does reading diffs. The only check that keeps meaning is the one a customer would perform: can somebody still sign up, still check out, still log back in tomorrow.

The trap is testing at the same level of care as the coding. A generated test suite, unreviewed, verifying generated code, unreviewed, is a closed loop that produces confidence and no information. The sentences describing what must work are the part a person should still own.

in smolanalytics

A test is a sentence you write and keep, so the part that encodes intent stays human even when the implementation does not. The suggest command walks your running application and drafts a suite from the flows it can find, and you delete the ones you disagree with; nothing it proposes is invented, because each one quotes what the walk actually found on the page.

How it works shows where this fits in the loop, and the docs have the exact behaviour.

questions

Is this just end-to-end testing with a new name?
Mostly yes, and the name matters anyway, because the failure mode is different. Traditional suites decay because nobody maintains them. Here the code changes faster than anybody reads it, so the tests are not competing with review, they are replacing it.

keep reading