The argument is in another post: splat captures encode real world light, so the engine's job is to show the photograph, not reinterpret it. That commitment becomes a series of small "no, not that" calls. Five of them, this build:
1. The invented atmosphere chip. PRD §4.5.1 originally wanted a top-right HUD pill reading "ROOM · 13:42 LOCAL · 22 °C · CLEAR." It was never the wall-clock time and never the room's actual weather — both fields were going to be made up per-arena. We dropped it. The brain reads invented timestamps over a real photograph as a tell. The atmosphere chip is gone from the HUD entirely.
2. Default-on post-process. The visual-preset slider could have shipped at BALANCED by default. Instead the default is null — OFF — and the post chain only spools up when the user opts into a preset. Bloom, vignette, fringing, tonemap remap all wait for a tap. Cinematic preset opt-in is the one path that activates the per-arena colour LUTs.
3. Accessibility palette tints that touch the canvas. The colour-blind palette settings remap the HUD swatch only — the bash-accent amber shifts toward yellow-blue / blue-cyan / orange per deficiency. The captured scene stays untouched. A renderer-side colour adjustment would lie to the player about what they are looking at; we couldn't sell that as accessibility.
4. The brake grind that wasn't. The driving channels pair tightly (see /devlog/three-channels) but the brake-punch FOV (D2) ships without an audible brake-grind synth. Most cars stopping in most arenas don't make the sound the brain expects from a "real" brake-grind. The slot is open. We chose to leave it open.
5. Sun flare across every arena. A9 lens flare is opt-in per outdoor arena: LUDLOW has a sun, LA NIGHT has a streetlight, ROOM has nothing. Indoor captures don't get a flare because there is no sun in a room. The flare overlay is also reduced-motion gated — vestibular-sensitive viewers don't want a wobbling bloom on a moving camera.
The flip side matters too. We did add layers — collision blur, whole-mix concussion LPF, sodium-tinted fog on LA NIGHT — but every one of those has the same defensible answer: that's what a real camera or a real listener does in that moment. Blur after an impact is the lens-element shake; the LPF is what ears actually do under concussion; fog on a sodium-lit street is the haze that's in the photograph at depth.
The test is short: would adding this layer make a real photo look *less* real, or *more* like what you'd see if your eye were there? The first answer is a no. The second is a yes.