The Mirror Test
Chrome finally mirrors its environment. The reflection blur schedule is now logarithmic and keyed to resolution, brushed metal streaks its reflections like actual brushed metal, and clearcoat reflects at its own polish instead of borrowing the base layer's.
A mirror has one job. Until this release, a roughness-0.05 metal in Lux did not do that job: reflections came back as washed-out suggestions of the environment, and no amount of material tweaking could sharpen them. The sphere above is the fix passing its audition, reflecting a four-quadrant test environment with every boundary crisp and every hue where the geometry says it should be.
Blame the schedule, not the blur
Reflections at different roughness levels come from pre-blurred copies of the environment, stacked as mip levels. Lux spaced those copies evenly: perfect mirror, then 0.25, then 0.5, and so on. The gap between “perfect mirror” and “0.25” turns out to be where every glossy material on earth lives, so anything shinier than satin sampled a blend of a sharp image and a heavy blur. The result was the classic double-exposure look, a sharp reflection wearing a frosted veil.
The fix is the industry-standard schedule: blur levels spaced logarithmically, with a dedicated near-mirror level and the spacing keyed to mip resolution, so a given mip size always carries the same roughness. The shader-side lookup is the exact inverse of the bake-side schedule, both anchored by a round-trip unit test, so the lobe you ask for is the lobe you get. The mip count pin on EnvironmentMap is gone in the process; the chain depth is now derived from the resolution, because a schedule with missing rungs is not a schedule. The resolution pin remains as the one honest quality dial.
Brushed metal, before and after
The anisotropy work from the last materials pass bent reflection vectors along the surface grain, and then nobody could see it, partly because of the contrast problem above, and partly because of a second bug this release flushed out: the brushed-metal BSDF compiled itself as a plastic. Its specular response was 4% of what it claimed. With both fixed, the pair above is the same sphere with anisotropy off and on: the streak smears the environment along the grain while keeping structure across it, which is the entire personality of brushed metal.
Clearcoat got the same honesty treatment. The coat layer used to sample the environment at the base layer’s roughness, so polished lacquer over rough metal reflected a rough world. It now samples at its own polish. Five behavioral gates pin all of this down, including one that checks the rim of a dark glass ball out-reflects its centre, because Fresnel says it must, and one that fails if anisotropy ever degenerates into plain extra blur.
The white-furnace energy tests stayed green through all of it, which is the difference between making reflections brighter and making them right.