We use Google Analytics to see which pages help people. No ads, no profiles. Privacy Policy
Compose, Kotlin, and multiplatform work — the mechanism under the symptom.
229 posts
Material's type scale gives fifteen named roles instead of a font-size parameter. Naming the role rather than the size is what makes typography consistent across a team and correct at large font scales.
Dynamic colour derives a full Material scheme from the user's wallpaper. It improves the feel of most apps and destroys brand recognition if applied indiscriminately, so the interesting work is deciding what stays fixed.
MaterialTheme provides three CompositionLocals and a handful of defaults. Understanding its actual shape explains how components pick up styling, why hardcoded values break dark mode, and what a custom theme has to replace.
Material 3's SearchBar transitions into a full-screen suggestion surface, which makes it a navigation concern. Chips come in four semantic variants, and the assist/filter/input/suggestion split matters.
Compose's date and time pickers are straightforward components hiding a genuine trap: the selected value is UTC-midnight epoch millis, and converting it naively shifts dates by one in most of the world.
Compose gives you progress indicators; it doesn't tell you when to show one. Duration decides: under 300ms show nothing, over a second show progress, and never let an indicator replace content that could be a skeleton.
NavigationBar, NavigationRail and the drawer are three presentations of the same destination list. Window size decides which, and NavigationSuiteScaffold does the switching — with rules about how many destinations belong there at all.
Dialogs and bottom sheets are composables that exist conditionally, not imperative calls. Modelling them as state fixes the rotation bug, the double-show bug, and the question of where the result goes.
Compose has two TextField APIs. The String + onValueChange pair has real problems with cursor position and asynchronous updates; the TextFieldState overload solves them and is now the recommended one.
Scoping a context pack by project type saved 8.2% per subagent, not the 40% I guessed. The real risk is a thin pack: a missing surface makes a generator invent.
A durable source of truth and an in-memory one give the same answers until a process dies. Deciding which is authoritative is a design act, not a default.
Checkboxes, radio buttons and switches encode different semantics — multiple choice, exclusive choice, and immediate effect. Choosing by appearance produces interfaces that mislead, and the accessibility story differs too.
