diff --git a/src/main.zig b/src/main.zig
index a6f641a..8731ce0 100644
--- a/src/main.zig
+++ b/src/main.zig
@@ -2,10 +2,18 @@ const std = @import("std");
 const core = @import("mach-core");
 const gpu = core.gpu;
 
+const zm = @import("zmath");
+const vec = zm.f32x4;
+const Mat = zm.Mat;
+
 pub const App = @This();
 
+app_timer: core.Timer,
 title_timer: core.Timer,
+
 pipeline: *gpu.RenderPipeline,
+mvp_uniform_buffer: *gpu.Buffer,
+mvp_bind_group: *gpu.BindGroup,
 
 pub fn init(app: *App) !void {
     try core.init(.{});
@@ -32,13 +40,32 @@ pub fn init(app: *App) !void {
         .fragment = &fragment,
     };
 
+    app.app_timer = try core.Timer.start();
     app.title_timer = try core.Timer.start();
+
     app.pipeline = core.device.createRenderPipeline(&pipeline_descriptor);
+
+    app.mvp_uniform_buffer = core.device.createBuffer(&.{
+        .usage = .{ .copy_dst = true, .uniform = true },
+        .size = @sizeOf(zm.Mat),
+        .mapped_at_creation = .false,
+    });
+
+    app.mvp_bind_group = core.device.createBindGroup(
+        &gpu.BindGroup.Descriptor.init(.{
+            .layout = app.pipeline.getBindGroupLayout(0),
+            .entries = &.{
+                gpu.BindGroup.Entry.buffer(0, app.mvp_uniform_buffer, 0, @sizeOf(zm.Mat)),
+            },
+        }),
+    );
 }
 
 pub fn deinit(app: *App) void {
     defer core.deinit();
     defer app.pipeline.release();
+    defer app.mvp_uniform_buffer.release();
+    defer app.mvp_bind_group.release();
 }
 
 pub fn update(app: *App) !bool {
@@ -50,6 +77,24 @@ pub fn update(app: *App) !bool {
         }
     }
 
+    // Set up a view matrix from the camera transform.
+    // This moves everything to be relative to the camera.
+    // TODO: Actually implement camera transform instead of hardcoding a look-at matrix.
+    // const view_matrix = zm.inverse(app.camera_transform);
+    const time = app.app_timer.read();
+    const x = @cos(time * std.math.tau / 10);
+    const y = @sin(time * std.math.tau / 10);
+    const view_matrix = zm.lookAtLh(vec(x, y, -2, 1), vec(0, 0, 0, 1), vec(0, 1, 0, 1));
+
+    // Set up a projection matrix using the size of the window.
+    // The perspective projection will make things further away appear smaller.
+    const width: f32 = @floatFromInt(core.descriptor.width);
+    const height: f32 = @floatFromInt(core.descriptor.height);
+    const field_of_view = std.math.degreesToRadians(f32, 45.0);
+    const proj_matrix = zm.perspectiveFovLh(field_of_view, width / height, 0.1, 10);
+
+    const view_proj_matrix = zm.mul(view_matrix, proj_matrix);
+
     // Get back buffer texture to render to.
     const back_buffer_view = core.swap_chain.getCurrentTextureView().?;
     defer back_buffer_view.release();
@@ -70,15 +115,18 @@ pub fn update(app: *App) !bool {
     const encoder = core.device.createCommandEncoder(null);
     defer encoder.release();
 
+    encoder.writeBuffer(app.mvp_uniform_buffer, 0, &[_]zm.Mat{zm.transpose(view_proj_matrix)});
+
     {
         const pass = encoder.beginRenderPass(&render_pass_info);
         defer pass.release();
         defer pass.end();
 
         pass.setPipeline(app.pipeline);
+        pass.setBindGroup(0, app.mvp_bind_group, &.{});
 
-        // Draw a triangle with the help of a specialized shader.
-        pass.draw(3, 1, 0, 0);
+        // Draw three triangles with the help of a specialized shader.
+        pass.draw(9, 1, 0, 0);
     }
 
     // Finish recording commands, creating a `WGPUCommandBuffer`.
diff --git a/src/shader.wgsl b/src/shader.wgsl
index f34d2e2..34bf84e 100644
--- a/src/shader.wgsl
+++ b/src/shader.wgsl
@@ -1,14 +1,44 @@
+@group(0) @binding(0) var<uniform> mvp_matrix: mat4x4<f32>;
+
+struct VertexOutput {
+    @builtin(position) position: vec4<f32>,
+    @location(0) color: vec4<f32>,
+};
+
+struct FragmentOutput {
+    @location(0) pixel_color: vec4<f32>
+};
+
 @vertex fn vertex_main(
     @builtin(vertex_index) index: u32
-) -> @builtin(position) vec4<f32> {
-    let pos = array<vec2<f32>, 3>(
-        vec2<f32>( 0.0,  0.5),
-        vec2<f32>(-0.5, -0.5),
-        vec2<f32>( 0.5, -0.5)
+) -> VertexOutput {
+    let color = array<vec3<f32>, 3>(
+        vec3<f32>(1.0, 0.0, 0.0),
+        vec3<f32>(0.0, 1.0, 0.0),
+        vec3<f32>(0.0, 0.0, 1.0),
     );
-    return vec4<f32>(pos[index], 0.0, 1.0);
+    let pos = array<vec2<f32>, 9>(
+        vec2<f32>(-1.0,  0.5 + 0.25),
+        vec2<f32>(-1.5, -0.5 + 0.25),
+        vec2<f32>(-0.5, -0.5 + 0.25),
+
+        vec2<f32>( 0.0,  0.5 - 0.25),
+        vec2<f32>(-0.5, -0.5 - 0.25),
+        vec2<f32>( 0.5, -0.5 - 0.25),
+
+        vec2<f32>( 1.0,  0.5),
+        vec2<f32>( 0.5, -0.5),
+        vec2<f32>( 1.5, -0.5),
+    );
+
+    var out: VertexOutput;
+    out.position = vec4<f32>(pos[index], 0.0, 1.0) * mvp_matrix;
+    out.color    = vec4<f32>(color[index / 3], 1.0);
+    return out;
 }
 
-@fragment fn frag_main() -> @location(0) vec4<f32> {
-    return vec4<f32>(1.0, 0.0, 0.0, 1.0);
+@fragment fn frag_main(in: VertexOutput) -> FragmentOutput {
+    var out: FragmentOutput;
+    out.pixel_color = in.color;
+    return out;
 }