I spent a lot of time on details that are easy to underestimate.
Scroll that does not fight the user. A keyboard that gets out of the way. A task edit sheet that feels stable. Drag behavior that follows your finger. Controls that are close enough to reach, but not so loud that the screen turns into cockpit soup.
None of that sounds like the glamorous part of building an AI product.
But for Nala, it is the product.
If the interface fights your hands, the assistant loses trust before the AI even starts working.
An AI task manager has to feel calm first.
A normal TODO app can get away with a surprising amount of friction. It stores the task. It shows the list. It checks the box.
Nala is being built toward something heavier: an AI personal assistant that can understand work, keep context, and help move tasks forward — including routing work to agents when that makes sense.
That means the app cannot feel like a spreadsheet with notifications.
Before the assistant can feel smart, the basic movement has to feel right. Opening a task. Editing it. Changing time. Choosing a date. Dismissing the keyboard. Dragging the sheet down. Scrolling through a conversation without the app yanking the screen back like a nervous intern.
These are small interactions, but they decide whether the product feels safe.
Scroll should not argue.
When the user scrolls manually, the app should respect that. Auto-scroll is useful until it becomes a tug-of-war.
The keyboard is part of the UI.
On mobile, the keyboard changes the entire screen. If it traps focus or covers work, the task suddenly feels heavier.
The sheet should feel physical.
A task edit modal is not just a panel. It needs stable presentation, a clear backdrop, and drag behavior that feels obvious.
Controls need hierarchy.
Priority, time, due date, business context, delegation, and Done all matter — but not all at the same volume.
Perfect UX usually hides inside boring fixes.
The work from today was not one huge shiny feature. It was a pile of tiny corrections around task editing, chat behavior, and the paths people use without thinking.
Stop chat auto-scroll from fighting manual scrolling.
Dismiss the task edit keyboard when tapping the background.
Stabilize the task edit modal. Tune the drag. Clarify the controls. Move delegation where it belongs. Make the backdrop feel intentional. Remove little visual artifacts that make a screen feel cheap even when the logic is correct.
Then the same pass widened: notification taps needed to route to the right place, chat behavior needed to stay steady, calendar and Today screen details needed to feel less brittle, and memory editing needed to behave like part of the product instead of an admin corner.
None of those are headline features. All of them decide whether the app feels like it respects the user’s hand.
That kind of work is easy to skip because every individual fix looks small.
Together, it changes the character of the app.
The tiny stuff is not decoration.
In productivity software, friction compounds.
If editing a task feels annoying once, fine. If it feels annoying ten times a day, the product starts charging rent in your brain.
For an AI task manager, that is even more important. The user is not only storing data. They are handing the assistant pieces of their workday. A task might become a reminder, a calendar item, a note, a client action, or eventually a handoff to an AI agent.
So the editing surface has to feel clean enough for simple tasks, but structured enough for real work.
- The app should know when to move the screen — and when to leave it alone.
- The keyboard should appear and disappear without turning the layout into a magic trick.
- The edit sheet should feel like an object you can control, not a popup that happened to you.
- The main action should be obvious without making every secondary action disappear.
- Delegation should feel like part of task flow, not an advanced setting hidden in a basement.
Good AI UX starts before the AI.
There is a trap in building AI products: treating the model as the only thing that matters.
A stronger model helps. Better memory helps. Agent routing helps. But the user experiences all of that through touch, scroll, typing, waiting, editing, and tiny moments of feedback.
If those moments feel unstable, the product feels unstable.
That is why I care so much about the details. Not because every pixel needs ceremony. Because an assistant is only useful if people trust it enough to keep putting work into it.
Trust is not created by a headline feature. Trust is built when the product quietly does the right thing again and again.
The product lesson
The better Nala gets, the more important the quiet parts become.
- Make task editing feel calm.
- Respect the user’s hand.
- Keep powerful controls understandable.
- Turn AI into a natural next step, not another panel to manage.
- Remove every tiny fight before it becomes product debt.
Perfect UX is not one big moment. It is a thousand small decisions that make the app feel like it is on your side.
Quick answers
Why does UX matter so much in an AI task manager?
Because the user is trusting the product with real work. If the basics feel jumpy or heavy, the assistant feels less reliable even before the AI features are involved.
What tiny UX details matter most here?
Scroll behavior, keyboard dismissal, edit-sheet stability, drag gestures, control hierarchy, saved state, due dates, time estimates, and delegation controls.
Is this a launch announcement?
No. Nala is still pre-launch. This is a build-in-public product note about the UX direction and the kind of iteration going into the app.