diff --git a/README.md b/README.md
index b4be02e..2db47e0 100644
--- a/README.md
+++ b/README.md
@@ -1361,6 +1361,20 @@ Things I chose while building, with the reasoning:
     phase-1 suite (43 checks, including the real route mapping against a stub Ollama returning a
     monthly-limit body, a two-device sync round-trip, and the generation guards below).
 
+61. **Glass that sits still doesn't blur.** The MacBook got hot just browsing: the ice scene animates
+    constantly, and every `backdrop-filter` surface re-runs its blur on every frame while the backdrop
+    moves — the whole card and rail area was re-blurring the scene ~60 times a second. Measured on
+    desktop (2x CPU throttle, software renderer): p50 frame 66ms with everything blurring, 33ms once
+    the static surfaces stopped, 17ms with the scene's animation also off — frames over 34ms dropped
+    from 128 to 14. The recipe is now split: **floating glass** (sheets, dropdowns, dialogs, the
+    sticky header/chrome/tab bar — the layers that can have content scroll underneath) keeps the
+    blur, now `backdrop-blur-lg` (which is what phones already used), while **static glass** (cards,
+    the rail, the gate, the study prompt card) keeps the exact same tint, rim, highlight and shadow
+    without the filter. A deterministic before/after pixel diff (scene motion frozen) shows the
+    change is a ~1% colour shift in the card backdrops — invisible in both themes.
+    `tests/bubble-perf.mjs` now reports the glass cost too, so a future blurred surface over the
+    scene shows up immediately.
+
 60. **There is exactly one loading screen.** The deck wizard had the good one — the app icon, the
     rotating stage message as the headline, the progress bar, the stage checklist, a tip and
     Cancel — while the homework flows showed a spinner with a title and the route loads a pulsing
diff --git a/app/globals.css b/app/globals.css
index 90a0b50..4357eb8 100644
--- a/app/globals.css
+++ b/app/globals.css
@@ -404,8 +404,20 @@ body:has(.overlay-open) :is(.app-header, .segmented, .tabbar) {
 
 @layer components {
 /* Frosted ice glass ------------------------------------------------------
-   One shared recipe for the large surfaces only: translucent tint, 24px
-   blur, a bright rim, an inner top highlight and a blue-tinted shadow.
+   Two recipes now:
+
+   - **Floating glass** (sheets, dropdowns, dialogs and the sticky chrome):
+     the tint below plus a backdrop blur — the blur is what reads as glass
+     when content scrolls underneath (the app header, the study chrome, the
+     tab bar, a sheet over the page).
+   - **Static glass** (cards, the rail, the gate, the study prompt card): the
+     exact same tint, rim, highlight and shadow, but **no** backdrop-filter.
+     On a smooth gradient the blur is invisible — and because the ice scene
+     animates, every blurred surface re-runs the filter on every frame.
+     Measured on desktop (2x throttle, software renderer): p50 frame time
+     66ms while the static surfaces blurred the scene, 33ms without them,
+     17ms with the scene's animation also stopped.
+
    Nested glass (chips, inputs) never adds a second blur layer. Tailwind's
    backdrop utilities emit the -webkit- prefix too (hand-written
    backdrop-filter gets minified to -webkit- only, which paints nothing). */
@@ -416,11 +428,18 @@ body:has(.overlay-open) :is(.app-header, .segmented, .tabbar) {
 .tabbar,
 .glass,
 .gate-card {
-  @apply backdrop-blur-xl backdrop-saturate-[1.4];
   background: var(--glass-bg);
   box-shadow: var(--glass-shadow), inset 0 1px 0 var(--glass-highlight);
 }
 
+/* Only the layers that can have content behind them blur. */
+.glass-panel,
+.study-chrome,
+.tabbar,
+.glass {
+  @apply backdrop-blur-lg backdrop-saturate-[1.4];
+}
+
 .glass-panel {
   border: 1px solid var(--glass-border);
   border-radius: 1.5rem;
@@ -460,28 +479,23 @@ body:has(.overlay-open) :is(.app-header, .segmented, .tabbar) {
   border-radius: 1.75rem;                /* 28px */
 }
 
-/* Blur is expensive on phones: halve it, and thin out the ice scene. */
+/* Blur is expensive: halve it on phones (floating glass only). */
 @media (max-width: 1023px) {
   .glass-panel,
-  .card,
-  .sidebar-surface,
   .study-chrome,
   .tabbar,
-  .glass,
-  .gate-card {
-    @apply backdrop-blur-lg;
+  .glass {
+    @apply backdrop-blur-md;
   }
 }
 
-/* No backdrop-filter support: fall back to a more opaque surface. */
+/* No backdrop-filter support: fall back to a more opaque surface (floating
+   glass only — the static surfaces already look right without a filter). */
 @supports not ((backdrop-filter: blur(1px)) or (-webkit-backdrop-filter: blur(1px))) {
   .glass-panel,
-  .card,
-  .sidebar-surface,
   .study-chrome,
   .tabbar,
-  .glass,
-  .gate-card {
+  .glass {
     background: var(--glass-fallback);
   }
 }
diff --git a/components/activities/shared.tsx b/components/activities/shared.tsx
index 4185cd9..9c520fd 100644
--- a/components/activities/shared.tsx
+++ b/components/activities/shared.tsx
@@ -78,7 +78,7 @@ export function PromptCard({
 }) {
   return (
     <div
-      className={cn("glass-panel glass-solid rounded-[1.5rem] px-5 py-6 text-balance sm:px-6", className)}
+      className={cn("card glass-solid rounded-[1.5rem] px-5 py-6 text-balance sm:px-6", className)}
       style={style}
     >
       {children}
diff --git a/public/sw.js b/public/sw.js
index b083a25..6b8d422 100644
--- a/public/sw.js
+++ b/public/sw.js
@@ -9,7 +9,7 @@
  *
  * Bumping CACHE_VERSION is how a new deploy invalidates old assets.
  */
-const CACHE_VERSION = "freezmo-v48";
+const CACHE_VERSION = "freezmo-v49";
 const SHELL_CACHE = `${CACHE_VERSION}-shell`;
 const ASSET_CACHE = `${CACHE_VERSION}-assets`;
 
diff --git a/tests/bubble-perf.mjs b/tests/bubble-perf.mjs
index 94953e0..1888e2a 100644
--- a/tests/bubble-perf.mjs
+++ b/tests/bubble-perf.mjs
@@ -14,7 +14,7 @@ import { BASE, DECK, launchBrowser, makePage, seed } from "./lib.mjs";
 const label = process.argv[2] ?? "run";
 const browser = await launchBrowser();
 
-async function measure(name, { width, height, throttle, reducedMotion }) {
+async function measure(name, { width, height, throttle, reducedMotion, extraCss }) {
   const { context, page } = await makePage(browser, { width, height, deviceScaleFactor: 2, reducedMotion });
   const cdp = await context.newCDPSession(page);
   await cdp.send("Emulation.setCPUThrottlingRate", { rate: throttle });
@@ -22,6 +22,10 @@ async function measure(name, { width, height, throttle, reducedMotion }) {
   await seed(page, { decks: [DECK] });
   await page.goto(BASE, { waitUntil: "networkidle" });
   await page.waitForTimeout(1500);
+  if (extraCss) {
+    await page.addStyleTag({ content: extraCss });
+    await page.waitForTimeout(400);
+  }
 
   const scene = await page.evaluate(() => {
     const orb = document.querySelector(".ice-orb-1");
@@ -79,6 +83,19 @@ const phone = { width: 390, height: 844 };
 const animated = await measure("phone 390 animated", { ...phone, throttle: 4 });
 const still = await measure("phone 390 reduced-motion", { ...phone, throttle: 4, reducedMotion: "reduce" });
 const desktop = await measure("desktop 1280 animated", { width: 1280, height: 900, throttle: 2 });
+/* The glass cost: how much of the frame time is backdrop-filter re-blurring
+   the (always animating) scene. Static surfaces no longer blur at all — see
+   the "Frosted ice glass" block in app/globals.css — so this number should
+   stay close to the animated run; a jump means a new blurred surface landed
+   on top of the scene. */
+const noGlass = await measure("desktop 1280 no backdrop-filter", {
+  width: 1280,
+  height: 900,
+  throttle: 2,
+  extraCss: "*{backdrop-filter:none !important;-webkit-backdrop-filter:none !important;}",
+});
 
-console.log(`\n[${label}] deltas — animated vs reduced-motion: p95 ${(animated.p95 - still.p95).toFixed(1)}ms, busy ${(animated.busyMs - still.busyMs).toFixed(0)}ms; desktop p95 ${desktop.p95.toFixed(1)}ms`);
+console.log(
+  `\n[${label}] deltas — animated vs reduced-motion: p95 ${(animated.p95 - still.p95).toFixed(1)}ms, busy ${(animated.busyMs - still.busyMs).toFixed(0)}ms; desktop p95 ${desktop.p95.toFixed(1)}ms; glass cost on desktop: p95 ${(desktop.p95 - noGlass.p95).toFixed(1)}ms, over-34ms frames ${desktop.over34 - noGlass.over34}`,
+);
 await browser.close();