App stores trained us to expect an icon for every job. Generative UI demos now suggest something else: describe the task, get a screen invented on the spot — forms, dashboards, flows assembled by a model.
So do we still need apps if AI can build the interface on demand?
What on-demand UI is good for
One-off internal tools. Personal views of the same data. Experiments that would never justify a sprint. Prototypes that update as the brief changes.
What durable apps still win
Reliability under load. Accessibility that was tested, not hoped for. Offline behaviour. Permissions, audit logs, brand consistency. Performance on low-end phones. A generated layout that reshuffles every session is a poor place to put payroll or medicine.
The likely middle
Stable shells — identity, navigation, data contracts — with AI filling disposable panels inside them. Or AI that proposes UI that engineers then harden into a real product.
We will need fewer mediocre apps. We will still need intentional ones. On-demand interfaces are a sharp knife for exploration; production trust still prefers something you can version, test, and blame when it breaks.
Brak komentarzy:
Prześlij komentarz