(a) THE MOVE

People open apps to get things done, not to hunt through messy navigation.

Capture the user's intent, open a tailored micro-UI, and let them finish in one tap. This app lets you build these experiences using AI models like Jev that map simple text directly to the right action.

Powered by Jev · a System One model
›
one field · the manifest decides what can happen

(b) THE PROBLEM

Designing an app where users never have to hunt through menus is incredibly difficult.

Instead of forcing users to climb through nested screens to find a single button, we’ve collapsed the workflow into a single field and panel. Because it accepts natural language, this exact pattern seamlessly adapts to voice and chat interfaces.

Before · nested menus

Settings
Reminders
Follow-ups
New follow-up

4 taps · 3 screens

After · one field

› schedule a follow-up
[ Follow-up panel ]

1 intent · 1 panel · 1 save

(c) THE MANIFEST

Define every intent once.

The developer writes a ledger of intents. That's the whole authoring surface.

intents.manifest
schedule a follow-up→ follow_up · tasks.db
log a work session→ session · log.db
capture a quick note→ note · notes.db
set a focus timer→ timer · tasks.db

(d) THE BUILD LOOP

The LLM runs once. Then Jev takes over as the "controller".

One LLM pass generates each micro UI and wires it to its persistence layer. After that, Jev is the controller, and each intent goes straight to its action.

1Define every intent in the manifest
2LLM generates the micro UI + persistence, once
3Jev routes each intent → matching micro UI → data persists

(e) THE FLIP

Define the intents you want. The rest is done.

Tell MicroUI what your people do, and it builds the fast path into every action without you writing component trees or wiring. Vibe-friendly, for non-developers.

Start building →