Apple's focus engine, put.io's list
A focusable view lifts: scale, light-source shadow, parallax tilt. You get it by making a view focusable and lose it by drawing your own highlight.
The tv-native fill sits under the lift, it does not replace it.
Top-level navigation is a TabView, not a header. The bar floats at the top, hides while content scrolls, and takes focus on swipe up. Search is a tab, not a header button.
The focused tab is a light pill with dark text. That contradicts the tv-native "never a white-invert pill" rule on purpose: tv-native is tier 4 where put.io owns focus, tvOS is tier 2 where the platform does. Selected-but-unfocused is plain bright text, no fill.
Everything else matches the other TV surfaces: a header and a list of destinations. No poster wall.
Twelve real filenames beat six invented posters.
No mono on TV. Two faces only, GT America regular and medium, with tabular figures for numerics.
Canvas
1920 × 1080
safe area from --tv-overscan-*
Tab bar
62px · 30px
TabView, focused pill is system
Row title
36px
tv.text.body · GT America Rg
Row subline
32px
tv.text.caption · tabular figures
Radius
--radius-tv 12px
one radius on every TV tier
Focused fill
--component-bg-active
under the system lift, not instead of it
Lift
system
scale + shadow owned by tvOS
Icon column
44px yellow
Phosphor, every file type
Numerics
GT America
tabular, no mono on TV