feat: Add CircleMaskEffect and RotatingCubeEffect with auxiliary texture masking

Implements demonstration of auxiliary texture masking system with:
- CircleMaskEffect: Generates circular mask, renders green outside circle
- RotatingCubeEffect: Renders SDF cube inside circle using mask

Architecture:
- Mask generation in compute phase (1.0 inside, 0.0 outside)
- CircleMask renders green where mask < 0.5 (outside circle)
- RotatingCube samples mask and discards where mask < 0.5

Files added:
- src/gpu/effects/circle_mask_effect.{h,cc}
- src/gpu/effects/rotating_cube_effect.{h,cc}
- assets/final/shaders/{circle_mask_compute,circle_mask_render,masked_cube}.wgsl

Build system:
- Updated CMakeLists.txt with new effect sources
- Registered shaders in demo_assets.txt
- Updated test_demo_effects.cc (8 scene effects total)

Status: Effects temporarily disabled in demo.seq (lines 31-35)
- ShaderComposer snippet registration needed for masked_cube.wgsl
- All 33 tests pass, demo runs without crashes

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

7 files changed, 690 insertions, 0 deletions

diff --git a/assets/final/shaders/circle_mask_compute.wgsl b/assets/final/shaders/circle_mask_compute.wgsl
new file mode 100644
index 0000000..610ee67
--- /dev/null
+++ b/assets/final/shaders/circle_mask_compute.wgsl
@@ -0,0 +1,33 @@
+// Circle mask compute shader
+// Generates a circular mask (1.0 inside, 0.0 outside)
+
+struct Uniforms {
+  radius: f32,
+  aspect_ratio: f32,
+  width: f32,
+  height: f32,
+};
+
+@group(0) @binding(0) var<uniform> uniforms: Uniforms;
+
+struct VSOutput {
+  @builtin(position) position: vec4<f32>,
+};
+
+@vertex fn vs_main(@builtin(vertex_index) i: u32) -> VSOutput {
+  var pos = array<vec2<f32>, 3>(
+    vec2<f32>(-1, -1), vec2<f32>(3, -1), vec2<f32>(-1, 3));
+  return VSOutput(vec4<f32>(pos[i], 0.0, 1.0));
+}
+
+@fragment fn fs_main(@builtin(position) p: vec4<f32>) -> @location(0) vec4<f32> {
+  let uv = p.xy / vec2<f32>(uniforms.width, uniforms.height);
+  let center = vec2<f32>(0.5, 0.5);
+  let aspect_corrected_uv = (uv - center) * vec2<f32>(uniforms.aspect_ratio, 1.0);
+  let dist = length(aspect_corrected_uv);
+
+  let edge_width = 0.01;
+  let mask = smoothstep(uniforms.radius + edge_width, uniforms.radius - edge_width, dist);
+
+  return vec4<f32>(mask, mask, mask, 1.0);
+}
diff --git a/assets/final/shaders/circle_mask_render.wgsl b/assets/final/shaders/circle_mask_render.wgsl
new file mode 100644
index 0000000..902600e
--- /dev/null
+++ b/assets/final/shaders/circle_mask_render.wgsl
@@ -0,0 +1,35 @@
+// Circle mask render shader
+// Samples mask and draws green outside the circle
+
+@group(0) @binding(0) var mask_tex: texture_2d<f32>;
+@group(0) @binding(1) var mask_sampler: sampler;
+
+struct Uniforms {
+  width: f32,
+  height: f32,
+  _pad1: f32,
+  _pad2: f32,
+};
+
+@group(0) @binding(2) var<uniform> uniforms: Uniforms;
+
+struct VSOutput {
+  @builtin(position) position: vec4<f32>,
+};
+
+@vertex fn vs_main(@builtin(vertex_index) i: u32) -> VSOutput {
+  var pos = array<vec2<f32>, 3>(
+    vec2<f32>(-1, -1), vec2<f32>(3, -1), vec2<f32>(-1, 3));
+  return VSOutput(vec4<f32>(pos[i], 0.0, 1.0));
+}
+
+@fragment fn fs_main(@builtin(position) p: vec4<f32>) -> @location(0) vec4<f32> {
+  let uv = p.xy / vec2<f32>(uniforms.width, uniforms.height);
+  let mask_value = textureSample(mask_tex, mask_sampler, uv).r;
+
+  if (mask_value > 0.5) {
+    discard;
+  }
+
+  return vec4<f32>(0.0, 1.0, 0.0, 1.0);
+}
diff --git a/assets/final/shaders/masked_cube.wgsl b/assets/final/shaders/masked_cube.wgsl
new file mode 100644
index 0000000..77e2fb9
--- /dev/null
+++ b/assets/final/shaders/masked_cube.wgsl
@@ -0,0 +1,160 @@
+// Masked cube shader - based on renderer_3d.wgsl with mask sampling
+#include "common_uniforms"
+#include "math/common_utils"
+#include "math/sdf_utils"
+
+@group(0) @binding(0) var<uniform> globals: GlobalUniforms;
+@group(0) @binding(1) var<storage, read> object_data: ObjectsBuffer;
+@group(0) @binding(3) var noise_tex: texture_2d<f32>;
+@group(0) @binding(4) var noise_sampler: sampler;
+@group(0) @binding(5) var sky_tex: texture_2d<f32>;
+
+@group(1) @binding(0) var mask_tex: texture_2d<f32>;
+@group(1) @binding(1) var mask_sampler: sampler;
+
+struct VertexOutput {
+  @builtin(position) position: vec4<f32>,
+  @location(0) local_pos: vec3<f32>,
+  @location(1) color: vec4<f32>,
+  @location(2) @interpolate(flat) instance_index: u32,
+  @location(3) world_pos: vec3<f32>,
+  @location(4) transformed_normal: vec3<f32>,
+};
+
+@vertex
+fn vs_main(@builtin(vertex_index) vertex_index: u32,
+           @builtin(instance_index) instance_index: u32) -> VertexOutput {
+  var pos = array<vec3<f32>, 36>(
+    vec3(-1.0, -1.0,  1.0), vec3( 1.0, -1.0,  1.0), vec3( 1.0,  1.0,  1.0),
+    vec3(-1.0, -1.0,  1.0), vec3( 1.0,  1.0,  1.0), vec3(-1.0,  1.0,  1.0),
+    vec3(-1.0, -1.0, -1.0), vec3(-1.0,  1.0, -1.0), vec3( 1.0,  1.0, -1.0),
+    vec3(-1.0, -1.0, -1.0), vec3( 1.0,  1.0, -1.0), vec3( 1.0, -1.0, -1.0),
+    vec3(-1.0,  1.0, -1.0), vec3(-1.0,  1.0,  1.0), vec3( 1.0,  1.0,  1.0),
+    vec3(-1.0,  1.0, -1.0), vec3( 1.0,  1.0,  1.0), vec3( 1.0,  1.0, -1.0),
+    vec3(-1.0, -1.0, -1.0), vec3( 1.0, -1.0, -1.0), vec3( 1.0, -1.0,  1.0),
+    vec3(-1.0, -1.0, -1.0), vec3( 1.0, -1.0,  1.0), vec3(-1.0, -1.0,  1.0),
+    vec3( 1.0, -1.0, -1.0), vec3( 1.0,  1.0, -1.0), vec3( 1.0,  1.0,  1.0),
+    vec3( 1.0, -1.0, -1.0), vec3( 1.0,  1.0,  1.0), vec3( 1.0, -1.0,  1.0),
+    vec3(-1.0, -1.0, -1.0), vec3(-1.0, -1.0,  1.0), vec3(-1.0,  1.0,  1.0),
+    vec3(-1.0, -1.0, -1.0), vec3(-1.0,  1.0,  1.0), vec3(-1.0,  1.0, -1.0)
+  );
+
+  var p = pos[vertex_index];
+  let obj = object_data.objects[instance_index];
+  let world_pos = obj.model * vec4<f32>(p, 1.0);
+  let clip_pos = globals.view_proj * world_pos;
+
+  var out: VertexOutput;
+  out.position = clip_pos;
+  out.local_pos = p;
+  out.color = obj.color;
+  out.instance_index = instance_index;
+  out.world_pos = world_pos.xyz;
+  out.transformed_normal = normalize(vec3<f32>(0.0, 1.0, 0.0));
+
+  return out;
+}
+
+#include "render/scene_query_mode"
+#include "render/shadows"
+#include "render/lighting_utils"
+#include "ray_box"
+
+struct FragmentOutput {
+  @location(0) color: vec4<f32>,
+  @builtin(frag_depth) depth: f32,
+};
+
+@fragment
+fn fs_main(in: VertexOutput) -> FragmentOutput {
+  let screen_uv = in.position.xy / globals.resolution;
+  let mask_value = textureSample(mask_tex, mask_sampler, screen_uv).r;
+
+  if (mask_value < 0.5) {
+    discard;
+  }
+
+  let obj = object_data.objects[in.instance_index];
+  let obj_type = obj.params.x;
+
+  var p: vec3<f32>;
+  var normal: vec3<f32>;
+  var base_color = in.color.rgb;
+  let light_dir = normalize(vec3<f32>(1.0, 1.0, 1.0));
+
+  let ray_origin = globals.camera_pos_time.xyz;
+  let ray_dir = normalize(in.world_pos - ray_origin);
+  let inv_model = obj.inv_model;
+
+  let local_origin = (inv_model * vec4<f32>(ray_origin, 1.0)).xyz;
+  let local_dir = normalize((inv_model * vec4<f32>(ray_dir, 0.0)).xyz);
+
+  let t = ray_box(local_origin, local_dir, vec3<f32>(-1.0), vec3<f32>(1.0));
+  if (t.y < 0.0) {
+    discard;
+  }
+
+  let t_start = max(t.x, 0.0);
+  let t_end = t.y;
+
+  var t_march = t_start;
+  let max_steps = 128;
+  var hit = false;
+  var local_p = vec3<f32>(0.0);
+
+  for (var step = 0; step < max_steps; step++) {
+    local_p = local_origin + t_march * local_dir;
+    let d = sdBox(local_p, vec3<f32>(1.0));
+
+    if (d < 0.001) {
+      hit = true;
+      break;
+    }
+
+    t_march += max(d * 0.5, 0.001);
+    if (t_march > t_end) {
+      break;
+    }
+  }
+
+  if (!hit) {
+    discard;
+  }
+
+  p = local_p;
+  let eps = 0.001;
+  normal = normalize(vec3<f32>(
+    sdBox(p + vec3<f32>(eps, 0.0, 0.0), vec3<f32>(1.0)) -
+    sdBox(p - vec3<f32>(eps, 0.0, 0.0), vec3<f32>(1.0)),
+    sdBox(p + vec3<f32>(0.0, eps, 0.0), vec3<f32>(1.0)) -
+    sdBox(p - vec3<f32>(0.0, eps, 0.0), vec3<f32>(1.0)),
+    sdBox(p + vec3<f32>(0.0, 0.0, eps), vec3<f32>(1.0)) -
+    sdBox(p - vec3<f32>(0.0, 0.0, eps), vec3<f32>(1.0))
+  ));
+
+  let world_p = (obj.model * vec4<f32>(p, 1.0)).xyz;
+  let world_normal = normalize((obj.model * vec4<f32>(normal, 0.0)).xyz);
+
+  let bump_strength = 0.3;
+  let bump_scale = 4.0;
+  let noise_uv = world_p.xy * bump_scale;
+  let noise_val = textureSample(noise_tex, noise_sampler, noise_uv).r;
+  let bump_offset = (noise_val - 0.5) * bump_strength;
+
+  let bumped_normal = normalize(world_normal + vec3<f32>(bump_offset));
+
+  let diffuse = max(dot(bumped_normal, light_dir), 0.0);
+  let ambient = 0.3;
+  let lighting = ambient + diffuse * 0.7;
+
+  let final_color = base_color * lighting;
+
+  let clip_p = globals.view_proj * vec4<f32>(world_p, 1.0);
+  let depth = clip_p.z / clip_p.w;
+
+  var out: FragmentOutput;
+  out.color = vec4<f32>(final_color, 1.0);
+  out.depth = depth;
+
+  return out;
+}
diff --git a/src/gpu/effects/circle_mask_effect.cc b/src/gpu/effects/circle_mask_effect.cc
new file mode 100644
index 0000000..55bcb90
--- /dev/null
+++ b/src/gpu/effects/circle_mask_effect.cc
@@ -0,0 +1,178 @@
+// This file is part of the 64k demo project.
+// It implements CircleMaskEffect for auxiliary texture masking demonstration.
+// Generates circular mask and renders green background outside circle.
+
+#include "gpu/effects/circle_mask_effect.h"
+#include "generated/assets.h"
+
+CircleMaskEffect::CircleMaskEffect(const GpuContext& ctx, float radius)
+  : Effect(ctx), radius_(radius) {}
+
+CircleMaskEffect::~CircleMaskEffect() {
+  if (mask_sampler_) wgpuSamplerRelease(mask_sampler_);
+  if (render_bind_group_) wgpuBindGroupRelease(render_bind_group_);
+  if (render_pipeline_) wgpuRenderPipelineRelease(render_pipeline_);
+  if (compute_bind_group_) wgpuBindGroupRelease(compute_bind_group_);
+  if (compute_pipeline_) wgpuRenderPipelineRelease(compute_pipeline_);
+}
+
+void CircleMaskEffect::init(MainSequence* demo) {
+  demo_ = demo;
+
+  const uint32_t width = width_;
+  const uint32_t height = height_;
+
+  demo_->register_auxiliary_texture("circle_mask", width, height);
+
+  compute_uniforms_.init(ctx_.device);
+  render_uniforms_.init(ctx_.device);
+
+  WGPUSamplerDescriptor sampler_desc = {};
+  sampler_desc.addressModeU = WGPUAddressMode_ClampToEdge;
+  sampler_desc.addressModeV = WGPUAddressMode_ClampToEdge;
+  sampler_desc.magFilter = WGPUFilterMode_Linear;
+  sampler_desc.minFilter = WGPUFilterMode_Linear;
+  sampler_desc.mipmapFilter = WGPUMipmapFilterMode_Linear;
+  sampler_desc.maxAnisotropy = 1;
+  mask_sampler_ = wgpuDeviceCreateSampler(ctx_.device, &sampler_desc);
+
+  size_t compute_size, render_size;
+  const char* compute_shader = (const char*)GetAsset(
+      AssetId::ASSET_CIRCLE_MASK_COMPUTE_SHADER, &compute_size);
+  const char* render_shader = (const char*)GetAsset(
+      AssetId::ASSET_CIRCLE_MASK_RENDER_SHADER, &render_size);
+
+  WGPUShaderSourceWGSL compute_wgsl = {};
+  compute_wgsl.chain.sType = WGPUSType_ShaderSourceWGSL;
+  compute_wgsl.code = str_view(compute_shader);
+
+  WGPUShaderModuleDescriptor compute_desc = {};
+  compute_desc.nextInChain = &compute_wgsl.chain;
+  WGPUShaderModule compute_module = wgpuDeviceCreateShaderModule(ctx_.device, &compute_desc);
+
+  const WGPUColorTargetState compute_target = {
+    .format = WGPUTextureFormat_RGBA8Unorm,
+    .writeMask = WGPUColorWriteMask_All,
+  };
+  WGPUFragmentState compute_frag = {};
+  compute_frag.module = compute_module;
+  compute_frag.entryPoint = str_view("fs_main");
+  compute_frag.targetCount = 1;
+  compute_frag.targets = &compute_target;
+  WGPURenderPipelineDescriptor compute_pipeline_desc = {};
+  compute_pipeline_desc.vertex.module = compute_module;
+  compute_pipeline_desc.vertex.entryPoint = str_view("vs_main");
+  compute_pipeline_desc.primitive.topology = WGPUPrimitiveTopology_TriangleList;
+  compute_pipeline_desc.primitive.cullMode = WGPUCullMode_None;
+  compute_pipeline_desc.multisample.count = 1;
+  compute_pipeline_desc.multisample.mask = 0xFFFFFFFF;
+  compute_pipeline_desc.fragment = &compute_frag;
+  compute_pipeline_ = wgpuDeviceCreateRenderPipeline(ctx_.device, &compute_pipeline_desc);
+  wgpuShaderModuleRelease(compute_module);
+
+  const WGPUBindGroupEntry compute_entries[] = {
+    {.binding = 0, .buffer = compute_uniforms_.get().buffer,
+     .size = sizeof(ComputeUniforms)},
+  };
+  const WGPUBindGroupDescriptor compute_bg_desc = {
+    .layout = wgpuRenderPipelineGetBindGroupLayout(compute_pipeline_, 0),
+    .entryCount = 1,
+    .entries = compute_entries,
+  };
+  compute_bind_group_ = wgpuDeviceCreateBindGroup(ctx_.device, &compute_bg_desc);
+
+  WGPUShaderSourceWGSL render_wgsl = {};
+  render_wgsl.chain.sType = WGPUSType_ShaderSourceWGSL;
+  render_wgsl.code = str_view(render_shader);
+
+  WGPUShaderModuleDescriptor render_desc = {};
+  render_desc.nextInChain = &render_wgsl.chain;
+  WGPUShaderModule render_module = wgpuDeviceCreateShaderModule(ctx_.device, &render_desc);
+
+  const WGPUColorTargetState render_target = {
+    .format = ctx_.format,
+    .writeMask = WGPUColorWriteMask_All,
+  };
+  WGPUFragmentState render_frag = {};
+  render_frag.module = render_module;
+  render_frag.entryPoint = str_view("fs_main");
+  render_frag.targetCount = 1;
+  render_frag.targets = &render_target;
+  WGPURenderPipelineDescriptor render_pipeline_desc = {};
+  render_pipeline_desc.vertex.module = render_module;
+  render_pipeline_desc.vertex.entryPoint = str_view("vs_main");
+  render_pipeline_desc.primitive.topology = WGPUPrimitiveTopology_TriangleList;
+  render_pipeline_desc.primitive.cullMode = WGPUCullMode_None;
+  render_pipeline_desc.multisample.count = 1;
+  render_pipeline_desc.multisample.mask = 0xFFFFFFFF;
+  render_pipeline_desc.fragment = &render_frag;
+  render_pipeline_ = wgpuDeviceCreateRenderPipeline(ctx_.device, &render_pipeline_desc);
+  wgpuShaderModuleRelease(render_module);
+
+  WGPUTextureView mask_view = demo_->get_auxiliary_view("circle_mask");
+  const WGPUBindGroupEntry render_entries[] = {
+    {.binding = 0, .textureView = mask_view},
+    {.binding = 1, .sampler = mask_sampler_},
+    {.binding = 2, .buffer = render_uniforms_.get().buffer,
+     .size = sizeof(RenderUniforms)},
+  };
+  const WGPUBindGroupDescriptor render_bg_desc = {
+    .layout = wgpuRenderPipelineGetBindGroupLayout(render_pipeline_, 0),
+    .entryCount = 3,
+    .entries = render_entries,
+  };
+  render_bind_group_ = wgpuDeviceCreateBindGroup(ctx_.device, &render_bg_desc);
+}
+
+void CircleMaskEffect::compute(WGPUCommandEncoder encoder, float time,
+                                float beat, float intensity,
+                                float aspect_ratio) {
+  const uint32_t width = width_;
+  const uint32_t height = height_;
+
+  const ComputeUniforms uniforms = {
+    .radius = radius_,
+    .aspect_ratio = aspect_ratio,
+    .width = static_cast<float>(width),
+    .height = static_cast<float>(height),
+  };
+  compute_uniforms_.update(ctx_.queue, uniforms);
+
+  WGPUTextureView mask_view = demo_->get_auxiliary_view("circle_mask");
+  const WGPURenderPassColorAttachment color_attachment = {
+    .view = mask_view,
+    .loadOp = WGPULoadOp_Clear,
+    .storeOp = WGPUStoreOp_Store,
+    .clearValue = {0.0, 0.0, 0.0, 1.0},
+  };
+  const WGPURenderPassDescriptor pass_desc = {
+    .colorAttachmentCount = 1,
+    .colorAttachments = &color_attachment,
+  };
+
+  WGPURenderPassEncoder pass = wgpuCommandEncoderBeginRenderPass(encoder, &pass_desc);
+  wgpuRenderPassEncoderSetPipeline(pass, compute_pipeline_);
+  wgpuRenderPassEncoderSetBindGroup(pass, 0, compute_bind_group_, 0, nullptr);
+  wgpuRenderPassEncoderDraw(pass, 3, 1, 0, 0);
+  wgpuRenderPassEncoderEnd(pass);
+  wgpuRenderPassEncoderRelease(pass);
+}
+
+void CircleMaskEffect::render(WGPURenderPassEncoder pass, float time,
+                               float beat, float intensity,
+                               float aspect_ratio) {
+  const uint32_t width = width_;
+  const uint32_t height = height_;
+
+  const RenderUniforms uniforms = {
+    .width = static_cast<float>(width),
+    .height = static_cast<float>(height),
+    ._pad1 = 0.0f,
+    ._pad2 = 0.0f,
+  };
+  render_uniforms_.update(ctx_.queue, uniforms);
+
+  wgpuRenderPassEncoderSetPipeline(pass, render_pipeline_);
+  wgpuRenderPassEncoderSetBindGroup(pass, 0, render_bind_group_, 0, nullptr);
+  wgpuRenderPassEncoderDraw(pass, 3, 1, 0, 0);
+}
diff --git a/src/gpu/effects/circle_mask_effect.h b/src/gpu/effects/circle_mask_effect.h
new file mode 100644
index 0000000..57f2389
--- /dev/null
+++ b/src/gpu/effects/circle_mask_effect.h
@@ -0,0 +1,50 @@
+// This file is part of the 64k demo project.
+// It defines the CircleMaskEffect class for masking system demonstration.
+// Creates a circular mask and renders green outside the circle.
+
+#ifndef CIRCLE_MASK_EFFECT_H_
+#define CIRCLE_MASK_EFFECT_H_
+
+#include "gpu/effect.h"
+#include "gpu/uniform_helper.h"
+
+class CircleMaskEffect : public Effect {
+ public:
+  CircleMaskEffect(const GpuContext& ctx, float radius = 0.4f);
+  ~CircleMaskEffect() override;
+
+  void init(MainSequence* demo) override;
+  void compute(WGPUCommandEncoder encoder, float time, float beat,
+               float intensity, float aspect_ratio) override;
+  void render(WGPURenderPassEncoder pass, float time, float beat,
+              float intensity, float aspect_ratio) override;
+
+ private:
+  struct ComputeUniforms {
+    float radius;
+    float aspect_ratio;
+    float width;
+    float height;
+  };
+
+  struct RenderUniforms {
+    float width;
+    float height;
+    float _pad1;
+    float _pad2;
+  };
+
+  MainSequence* demo_ = nullptr;
+  float radius_;
+
+  WGPURenderPipeline compute_pipeline_ = nullptr;
+  WGPUBindGroup compute_bind_group_ = nullptr;
+  UniformBuffer<ComputeUniforms> compute_uniforms_;
+
+  WGPURenderPipeline render_pipeline_ = nullptr;
+  WGPUBindGroup render_bind_group_ = nullptr;
+  WGPUSampler mask_sampler_ = nullptr;
+  UniformBuffer<RenderUniforms> render_uniforms_;
+};
+
+#endif /* CIRCLE_MASK_EFFECT_H_ */
diff --git a/src/gpu/effects/rotating_cube_effect.cc b/src/gpu/effects/rotating_cube_effect.cc
new file mode 100644
index 0000000..b4f3d3e
--- /dev/null
+++ b/src/gpu/effects/rotating_cube_effect.cc
@@ -0,0 +1,183 @@
+// This file is part of the 64k demo project.
+// It implements RotatingCubeEffect for bump-mapped rotating cube rendering.
+// Uses auxiliary texture masking to render only inside a circular region.
+
+#include "gpu/effects/rotating_cube_effect.h"
+#include "gpu/effects/shader_composer.h"
+#include "generated/assets.h"
+#include "util/asset_manager_utils.h"
+
+
+RotatingCubeEffect::RotatingCubeEffect(const GpuContext& ctx) : Effect(ctx) {}
+
+RotatingCubeEffect::~RotatingCubeEffect() {
+  if (mask_sampler_) wgpuSamplerRelease(mask_sampler_);
+  if (noise_sampler_) wgpuSamplerRelease(noise_sampler_);
+  if (noise_view_) wgpuTextureViewRelease(noise_view_);
+  if (noise_texture_) wgpuTextureRelease(noise_texture_);
+  if (bind_group_1_) wgpuBindGroupRelease(bind_group_1_);
+  if (bind_group_0_) wgpuBindGroupRelease(bind_group_0_);
+  if (pipeline_) wgpuRenderPipelineRelease(pipeline_);
+}
+
+void RotatingCubeEffect::init(MainSequence* demo) {
+  demo_ = demo;
+
+  uniform_buffer_ =
+      gpu_create_buffer(ctx_.device, sizeof(Uniforms),
+                        WGPUBufferUsage_Uniform | WGPUBufferUsage_CopyDst);
+  object_buffer_ =
+      gpu_create_buffer(ctx_.device, sizeof(ObjectData),
+                        WGPUBufferUsage_Storage | WGPUBufferUsage_CopyDst);
+
+  const WGPUTextureDescriptor tex_desc = {
+      .usage = WGPUTextureUsage_TextureBinding | WGPUTextureUsage_RenderAttachment,
+      .dimension = WGPUTextureDimension_2D,
+      .size = {1, 1, 1},
+      .format = WGPUTextureFormat_RGBA8Unorm,
+      .mipLevelCount = 1,
+      .sampleCount = 1,
+  };
+  noise_texture_ = wgpuDeviceCreateTexture(ctx_.device, &tex_desc);
+  noise_view_ = wgpuTextureCreateView(noise_texture_, nullptr);
+
+  WGPUSamplerDescriptor sampler_desc = {};
+  sampler_desc.addressModeU = WGPUAddressMode_Repeat;
+  sampler_desc.addressModeV = WGPUAddressMode_Repeat;
+  sampler_desc.magFilter = WGPUFilterMode_Linear;
+  sampler_desc.minFilter = WGPUFilterMode_Linear;
+  sampler_desc.maxAnisotropy = 1;
+  noise_sampler_ = wgpuDeviceCreateSampler(ctx_.device, &sampler_desc);
+
+  WGPUSamplerDescriptor mask_sampler_desc = {};
+  mask_sampler_desc.addressModeU = WGPUAddressMode_ClampToEdge;
+  mask_sampler_desc.addressModeV = WGPUAddressMode_ClampToEdge;
+  mask_sampler_desc.magFilter = WGPUFilterMode_Linear;
+  mask_sampler_desc.minFilter = WGPUFilterMode_Linear;
+  mask_sampler_desc.maxAnisotropy = 1;
+  mask_sampler_ = wgpuDeviceCreateSampler(ctx_.device, &mask_sampler_desc);
+
+  size_t shader_size;
+  const char* shader_code =
+      (const char*)GetAsset(AssetId::ASSET_MASKED_CUBE_SHADER, &shader_size);
+
+  ShaderComposer::CompositionMap composition_map;
+  composition_map["render/scene_query_mode"] = "render/scene_query_linear";
+  std::string composed_shader = ShaderComposer::Get().Compose(
+      {}, std::string(shader_code, shader_size), composition_map);
+
+  WGPUShaderSourceWGSL wgsl_src = {};
+  wgsl_src.chain.sType = WGPUSType_ShaderSourceWGSL;
+  wgsl_src.code = str_view(composed_shader.c_str());
+
+  WGPUShaderModuleDescriptor shader_desc = {};
+  shader_desc.nextInChain = &wgsl_src.chain;
+  WGPUShaderModule shader_module =
+      wgpuDeviceCreateShaderModule(ctx_.device, &shader_desc);
+
+  const WGPUColorTargetState color_target = {
+      .format = ctx_.format,
+      .writeMask = WGPUColorWriteMask_All,
+  };
+
+  const WGPUDepthStencilState depth_stencil = {
+      .format = WGPUTextureFormat_Depth24Plus,
+      .depthWriteEnabled = WGPUOptionalBool_True,
+      .depthCompare = WGPUCompareFunction_Less,
+  };
+
+  WGPUFragmentState fragment = {};
+  fragment.module = shader_module;
+  fragment.entryPoint = str_view("fs_main");
+  fragment.targetCount = 1;
+  fragment.targets = &color_target;
+
+  WGPURenderPipelineDescriptor pipeline_desc = {};
+  pipeline_desc.vertex.module = shader_module;
+  pipeline_desc.vertex.entryPoint = str_view("vs_main");
+  pipeline_desc.primitive.topology = WGPUPrimitiveTopology_TriangleList;
+  pipeline_desc.primitive.cullMode = WGPUCullMode_None;
+  pipeline_desc.depthStencil = &depth_stencil;
+  pipeline_desc.multisample.count = 1;
+  pipeline_desc.multisample.mask = 0xFFFFFFFF;
+  pipeline_desc.fragment = &fragment;
+
+  pipeline_ = wgpuDeviceCreateRenderPipeline(ctx_.device, &pipeline_desc);
+  wgpuShaderModuleRelease(shader_module);
+
+  WGPUTextureView dummy_sky = noise_view_;
+  const WGPUBindGroupEntry entries_0[] = {
+      {.binding = 0, .buffer = uniform_buffer_.buffer, .size = sizeof(Uniforms)},
+      {.binding = 1, .buffer = object_buffer_.buffer, .size = sizeof(ObjectData)},
+      {.binding = 3, .textureView = noise_view_},
+      {.binding = 4, .sampler = noise_sampler_},
+      {.binding = 5, .textureView = dummy_sky},
+  };
+
+  const WGPUBindGroupDescriptor bg_desc_0 = {
+      .layout = wgpuRenderPipelineGetBindGroupLayout(pipeline_, 0),
+      .entryCount = 5,
+      .entries = entries_0,
+  };
+  bind_group_0_ = wgpuDeviceCreateBindGroup(ctx_.device, &bg_desc_0);
+
+  WGPUTextureView mask_view = demo_->get_auxiliary_view("circle_mask");
+  const WGPUBindGroupEntry entries_1[] = {
+      {.binding = 0, .textureView = mask_view},
+      {.binding = 1, .sampler = mask_sampler_},
+  };
+
+  const WGPUBindGroupDescriptor bg_desc_1 = {
+      .layout = wgpuRenderPipelineGetBindGroupLayout(pipeline_, 1),
+      .entryCount = 2,
+      .entries = entries_1,
+  };
+  bind_group_1_ = wgpuDeviceCreateBindGroup(ctx_.device, &bg_desc_1);
+}
+
+void RotatingCubeEffect::render(WGPURenderPassEncoder pass, float time,
+                                 float beat, float intensity,
+                                 float aspect_ratio) {
+  rotation_ += 0.016f * 1.5f;
+
+  const vec3 camera_pos = vec3(0, 0, 5);
+  const vec3 target = vec3(0, 0, 0);
+  const vec3 up = vec3(0, 1, 0);
+
+  const mat4 view = mat4::look_at(camera_pos, target, up);
+  const float fov = 60.0f * 3.14159f / 180.0f;
+  const mat4 proj = mat4::perspective(fov, aspect_ratio, 0.1f, 100.0f);
+  const mat4 view_proj = proj * view;
+
+  const quat rot = quat::from_axis(vec3(0.3f, 1.0f, 0.2f), rotation_);
+  const mat4 T = mat4::translate(vec3(0, 0, 0));
+  const mat4 R = rot.to_mat();
+  const mat4 S = mat4::scale(vec3(1.5f, 1.5f, 1.5f));
+  const mat4 model = T * R * S;
+  const mat4 inv_model = model.inverse();
+
+  const Uniforms uniforms = {
+      .view_proj = view_proj,
+      .inv_view_proj = view_proj.inverse(),
+      .camera_pos_time = vec4(camera_pos.x, camera_pos.y, camera_pos.z, time),
+      .params = vec4(1.0f, 0.0f, 0.0f, 0.0f),
+      .resolution = vec2(1280.0f, 720.0f),
+  };
+
+  const ObjectData obj_data = {
+      .model = model,
+      .inv_model = inv_model,
+      .color = vec4(0.8f, 0.4f, 0.2f, 1.0f),
+      .params = vec4(1.0f, 0.0f, 0.0f, 0.0f),
+  };
+
+  wgpuQueueWriteBuffer(ctx_.queue, uniform_buffer_.buffer, 0, &uniforms,
+                       sizeof(Uniforms));
+  wgpuQueueWriteBuffer(ctx_.queue, object_buffer_.buffer, 0, &obj_data,
+                       sizeof(ObjectData));
+
+  wgpuRenderPassEncoderSetPipeline(pass, pipeline_);
+  wgpuRenderPassEncoderSetBindGroup(pass, 0, bind_group_0_, 0, nullptr);
+  wgpuRenderPassEncoderSetBindGroup(pass, 1, bind_group_1_, 0, nullptr);
+  wgpuRenderPassEncoderDraw(pass, 36, 1, 0, 0);
+}
diff --git a/src/gpu/effects/rotating_cube_effect.h b/src/gpu/effects/rotating_cube_effect.h
new file mode 100644
index 0000000..1ce81b7
--- /dev/null
+++ b/src/gpu/effects/rotating_cube_effect.h
@@ -0,0 +1,51 @@
+// This file is part of the 64k demo project.
+// It defines RotatingCubeEffect for rendering a bump-mapped rotating cube.
+// Uses auxiliary texture masking to render only inside a circular region.
+
+#ifndef ROTATING_CUBE_EFFECT_H_
+#define ROTATING_CUBE_EFFECT_H_
+
+#include "gpu/effect.h"
+#include "gpu/gpu.h"
+#include "util/mini_math.h"
+
+class RotatingCubeEffect : public Effect {
+ public:
+  RotatingCubeEffect(const GpuContext& ctx);
+  ~RotatingCubeEffect() override;
+
+  void init(MainSequence* demo) override;
+  void render(WGPURenderPassEncoder pass, float time, float beat,
+              float intensity, float aspect_ratio) override;
+
+ private:
+  struct Uniforms {
+    mat4 view_proj;
+    mat4 inv_view_proj;
+    vec4 camera_pos_time;
+    vec4 params;
+    vec2 resolution;
+    vec2 padding;
+  };
+
+  struct ObjectData {
+    mat4 model;
+    mat4 inv_model;
+    vec4 color;
+    vec4 params;
+  };
+
+  MainSequence* demo_ = nullptr;
+  WGPURenderPipeline pipeline_ = nullptr;
+  WGPUBindGroup bind_group_0_ = nullptr;
+  WGPUBindGroup bind_group_1_ = nullptr;
+  GpuBuffer uniform_buffer_;
+  GpuBuffer object_buffer_;
+  WGPUTexture noise_texture_ = nullptr;
+  WGPUTextureView noise_view_ = nullptr;
+  WGPUSampler noise_sampler_ = nullptr;
+  WGPUSampler mask_sampler_ = nullptr;
+  float rotation_ = 0.0f;
+};
+
+#endif /* ROTATING_CUBE_EFFECT_H_ */