From 017270f7e688e07a19b44d2bc237b07a65c258b2 Mon Sep 17 00:00:00 2001
From: Cameron <cameron@pfiffer.org>
Date: Thu, 9 Jul 2026 21:39:09 -0700
Subject: [PATCH] Bank the exposure drift downwind
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The even halo read as decoration; dust reads as accumulation. The
drift scatter now biases toward an airflow-shadow direction — fuller
reach downwind, thinned upwind, slight center offset — still a pure
deterministic function of the amount.

Defense: a [TUNE] pass within the decided exposure form
(mechanics/machine-work.md, amended with the uneven-settling note in
this commit): density and spread still carry the amount, paused
geometry still reads, no RNG or wall clock enters the renderer.
---
 src/bin/assets/rack.rs         | 22 +++++++++++++++++++---
 wiki/log/DEVLOG.md             | 10 ++++++++++
 wiki/mechanics/machine-work.md |  4 +++-
 3 files changed, 32 insertions(+), 4 deletions(-)

diff --git a/src/bin/assets/rack.rs b/src/bin/assets/rack.rs
index bc7da55..f1d534f 100644
--- a/src/bin/assets/rack.rs
+++ b/src/bin/assets/rack.rs
@@ -915,17 +915,33 @@ fn spawn_exposure_drift(
     let n = ((exposure * 16.0).ceil() as usize).min(140);
     let r0 = 0.52;
     let r1 = r0 + 0.16 * exposure.sqrt();
+    // Dust settles unevenly: the drift banks downwind of the machine
+    // (an even halo reads as decoration, not accumulation). BIAS is
+    // the airflow-shadow direction; upwind flecks thin out and the
+    // downwind bank reaches further.
+    const BIAS: f32 = 0.7; // radians, toward front-right
     for i in 0..n {
         let theta = i as f32 * 2.399_963; // golden angle
+        let downwind = (theta - BIAS).cos();
+        // Thin the upwind side: roughly two of three upwind flecks
+        // settle elsewhere.
+        if downwind < -0.25 && i % 3 != 0 {
+            continue;
+        }
         let frac = (i as f32 * 0.618_034) % 1.0;
-        let r = r0 + (r1 - r0) * frac.sqrt();
+        let reach = 1.0 + 0.45 * downwind.max(0.0);
+        let r = (r0 + (r1 - r0) * frac.sqrt()) * reach;
         let (sx, sz) = (theta.cos(), theta.sin() * 0.85);
         let s = if i % 3 == 0 { 0.030 } else { 0.018 };
         commands.spawn((
             Mesh3d(meshes.add(Cuboid::new(s, 0.008, s * 0.8))),
             MeshMaterial3d(fleck.clone()),
-            Transform::from_translation(Vec3::new(r * sx, 0.006, r * sz))
-                .with_rotation(Quat::from_rotation_y(i as f32 * 1.7)),
+            Transform::from_translation(Vec3::new(
+                r * sx + 0.06 * BIAS.cos(),
+                0.006,
+                r * sz + 0.05 * BIAS.sin(),
+            ))
+            .with_rotation(Quat::from_rotation_y(i as f32 * 1.7)),
             ChildOf(root),
         ));
     }
diff --git a/wiki/log/DEVLOG.md b/wiki/log/DEVLOG.md
index 20d4d34..94d8589 100644
--- a/wiki/log/DEVLOG.md
+++ b/wiki/log/DEVLOG.md
@@ -5,6 +5,16 @@ Type: log
 ```
 Reverse chronological implementation notes. Keep this factual: what changed, why, checks, and spec impact.
 
+## 2026-07-09 - Exposure drift banks downwind
+
+- Intent: the even drift halo read as decoration; dust should read as
+  accumulation.
+- Changed: the drift scatter biases downwind (fuller reach, thinned
+  upwind side, slight center offset), deterministic as before;
+  machine-work exposure bullet notes the uneven settling ([TUNE] bias).
+- Checks: full ./tools/check.sh; exposure board recapture.
+- Log: entry only (small [TUNE] pass on the decided form).
+
 ## 2026-07-09 - Exposure decided: particulate residue + footprint trails
 
 - Intent: close the exposure half of issue #1 — Cameron rejected the
diff --git a/wiki/mechanics/machine-work.md b/wiki/mechanics/machine-work.md
index 147756a..f004896 100644
--- a/wiki/mechanics/machine-work.md
+++ b/wiki/mechanics/machine-work.md
@@ -310,7 +310,9 @@ that can ride wires; exposure is contamination that cannot. The family kit is:
   exposure is dust"). Working machines shed faint crimson motes that settle
   into a **stippled drift** of fine angular flecks at the floor/base outside
   the amber ownership ring; density and spread are the amount, so a paused
-  frame carries the read. It absorbs light rather than emitting it — the one
+  frame carries the read. The drift settles unevenly — banked downwind, thin
+  upwind — so it reads as accumulation, never a decorative halo (bias
+  direction and thinning are [TUNE]). It absorbs light rather than emitting it — the one
   near-unlit material in a world of signals. Its physics and carrier roles
   stand: pulled by concealment wells (a vacuum draw on the dust), transfers
   onto people as attention, and materializes at the Office as filings
-- 
2.51.2