NN/g splits AI interfaces by who decided to build them, not who typed

Nielsen Norman Group's Amy Chang draws a line that's easy to blur: vibe coding is the AI executing a UI the user asked for, genUI is the AI deciding on its own that an interface is the right response. That distinction changes what you evaluate. A vibe-coded tool fails on fidelity to a stated request; genUI fails on judgment, meaning success has to be measured with user research and task metrics rather than "did it render."