diff --git a/build.zig b/build.zig
index 8b9ad378..1ca2b28d 100644
--- a/build.zig
+++ b/build.zig
@@ -52,6 +52,20 @@ pub fn build(b: *std.build.Builder) void {
         const example_run_step = b.step("run-example-" ++ example.name, "Run the example");
         example_run_step.dependOn(&example_run_cmd.step);
     }
+
+    const shaderexp_exe = b.addExecutable("shaderexp", "shaderexp/main.zig");
+    shaderexp_exe.setTarget(target);
+    shaderexp_exe.setBuildMode(mode);
+    shaderexp_exe.addPackage(pkg);
+    shaderexp_exe.addPackage(gpu.pkg);
+    shaderexp_exe.addPackage(glfw.pkg);
+    link(b, shaderexp_exe, options);
+    shaderexp_exe.install();
+
+    const shaderexp_run_cmd = shaderexp_exe.run();
+    shaderexp_run_cmd.step.dependOn(b.getInstallStep());
+    const shaderexp_run_step = b.step("run-shaderexp", "Run shaderexp");
+    shaderexp_run_step.dependOn(&shaderexp_run_cmd.step);
 }
 
 pub const Options = struct {
diff --git a/shaderexp/README.md b/shaderexp/README.md
new file mode 100644
index 00000000..e1beee4b
--- /dev/null
+++ b/shaderexp/README.md
@@ -0,0 +1,8 @@
+# Shaderexp
+
+This is an executable for testing wgsl shaders on the fly.
+Build it and run it with:
+`zig build run-shaderexp`
+
+Then modify shaderexp/frag.wgsl and save. The window will update and use the new fragment shader.
+If errors occur, the window will show a black_screen and an error message will be written to stdout.
diff --git a/shaderexp/black_screen_frag.wgsl b/shaderexp/black_screen_frag.wgsl
new file mode 100755
index 00000000..bc6f45d8
--- /dev/null
+++ b/shaderexp/black_screen_frag.wgsl
@@ -0,0 +1,11 @@
+struct UniformBufferObject {
+    time: f32,
+    resolution: vec2<f32>,
+}
+@group(0) @binding(0) var<uniform> ubo : UniformBufferObject;
+
+@stage(fragment) fn main(
+    @location(0) uv : vec2<f32>
+) -> @location(0) vec4<f32> {
+    return vec4<f32>( 0.0, 0.0, 0.0, 1.0);
+}
diff --git a/shaderexp/example-shaders/mandelbrot.wgsl b/shaderexp/example-shaders/mandelbrot.wgsl
new file mode 100755
index 00000000..c2058153
--- /dev/null
+++ b/shaderexp/example-shaders/mandelbrot.wgsl
@@ -0,0 +1,38 @@
+struct UniformBufferObject {
+    resolution: vec2<f32>,
+    time: f32,
+}
+@group(0) @binding(0) var<uniform> ubo : UniformBufferObject;
+
+@stage(fragment) fn main(
+    @location(0) uv : vec2<f32>
+) -> @location(0) vec4<f32> {
+    let aspect = ubo.resolution / min(ubo.resolution.x,ubo.resolution.y);
+    let translated_uv = (uv - vec2(0.5,0.5)) * aspect * 2.0;
+    let col = f32(mandel(translated_uv)) / 100.0;
+
+    return vec4(vec3<f32>(col), 1.0);
+}
+
+fn mandel(uv: vec2<f32>) -> i32{
+    let zoom = 1.0;
+    let center_position = vec2<f32>(0.5,0.0);
+    let mapped_point = uv * zoom - center_position;
+    var z = mapped_point;
+    var tmp:f32;
+    var i:i32 = 0;
+    var found = false;
+    var res = 0;
+    loop {
+        if (i >= 100){
+            break;
+        }
+        tmp = z.x;
+        z.x = z.x * z.x - z.y * z.y + mapped_point.x;
+        z.y = 2. * tmp * z.y + mapped_point.y;
+        found = found || (z.x * z.x + z.y * z.y > 16.);
+        res = res + 1 * i32(!found);
+        i = i + 1;
+    }
+    return res;
+}
diff --git a/shaderexp/example-shaders/ray_marching.wgsl b/shaderexp/example-shaders/ray_marching.wgsl
new file mode 100755
index 00000000..e783ec95
--- /dev/null
+++ b/shaderexp/example-shaders/ray_marching.wgsl
@@ -0,0 +1,98 @@
+// A slight modification / translation of https://www.shadertoy.com/view/XlGBW3
+// to WGSL.
+// License: Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License.
+
+struct UniformBufferObject {
+    resolution: vec2<f32>,
+    time: f32,
+}
+@group(0) @binding(0) var<uniform> ubo : UniformBufferObject;
+
+fn getDist(p:vec3<f32>) -> f32{
+    let dist_from_center:f32 = 2.*sin(ubo.time * 3.);
+    let rotation_speed:f32 = 6.;
+    
+    let sphere1 = vec4<f32>(dist_from_center*cos(rotation_speed*ubo.time + 3.14159 * 0. * 2. / 3.),1.1, dist_from_center*sin(rotation_speed*ubo.time + 3.14159 + 3.14159 * 0. * 2. / 3.),1.);
+    let sphere2 = vec4<f32>(dist_from_center*cos(rotation_speed*ubo.time + 3.14159 * 1. * 2. / 3.),1.1, dist_from_center*sin(rotation_speed*ubo.time + 3.14159 + 3.14159 * 1. * 2. / 3.),1.);
+    let sphere3 = vec4<f32>(dist_from_center*cos(rotation_speed*ubo.time + 3.14159 * 2. * 2. / 3.),1.1, dist_from_center*sin(rotation_speed*ubo.time + 3.14159 + 3.14159 * 2. * 2. / 3.),1.);
+
+    let sphere1_dist:f32 = length(p - sphere1.xyz) - sphere1.w;
+    let sphere2_dist:f32 = length(p - sphere2.xyz) - sphere2.w;
+    let sphere3_dist:f32 = length(p - sphere3.xyz) - sphere3.w;
+    let plane_dist = p.y;
+    
+    return min(min(min(sphere1_dist,sphere2_dist),sphere3_dist),plane_dist);
+}
+
+fn rayMarch(ro:vec3<f32>, rd:vec3<f32>) -> f32{
+    let MAX_STEPS:i32 = 100;
+    let MAX_DIST:f32 = 100.0;
+    let SURF_DIST:f32 = 0.01;
+    var d:f32 = 0.0;
+    
+    var i: i32 = 0;
+    loop {
+      if(i >= MAX_STEPS){
+        break;
+      }
+
+      let p = ro + rd * d;
+      let ds = getDist(p);
+      d = d + ds;
+      if(d > MAX_DIST || ds <= SURF_DIST){
+        break;
+      } 
+
+      i = i + 1;
+    }
+    return d;
+}
+
+fn getNormal(p:vec3<f32>) -> vec3<f32>{
+    let d = getDist(p);
+    let e = vec2<f32>(0.1,0.0);
+    
+    // We can find the normal using the points around the hit point
+    let n = d - vec3<f32>(
+        getDist(p-e.xyy),
+        getDist(p-e.yxy),
+        getDist(p-e.yyx)
+      );
+        
+    return normalize(n);
+}
+
+fn getLight(p:vec3<f32>) -> f32{
+    let SURF_DIST:f32 = .01;
+
+    let light_pos = vec3<f32>(0.,5.,0.);
+    let l = normalize(light_pos - p) * 1.;
+    let n = getNormal(p);
+    
+    var dif = clamp(dot(n,l),.0,1.);
+    
+    let d = rayMarch(p + n * SURF_DIST * 2.,l);
+    if(d<length(light_pos - p)){
+        dif = dif * .1;
+    }
+    
+    return dif;
+}
+
+@stage(fragment) fn main(
+    @location(0) uv : vec2<f32>
+) -> @location(0) vec4<f32> {
+    let aspect = ubo.resolution / min(ubo.resolution.x,ubo.resolution.y);
+    let tmp_uv = (uv - vec2(0.5,0.5)) * aspect * 2.0;
+    var col = vec3<f32>(0.0);
+    
+    let r_origin = vec3<f32>(4.0,3.,.0);
+    let r_dir = normalize(vec3<f32>(-1.0,tmp_uv.y,tmp_uv.x));
+    let d = rayMarch(r_origin,r_dir);
+    col = vec3<f32>(d / 8.);
+    let p = r_origin + r_dir * d;
+    let diff = getLight(p);
+    
+    col = vec3<f32>(diff , 0.,0.);
+    return vec4<f32>(col,0.0);
+}
diff --git a/shaderexp/example-shaders/test_shader.wgsl b/shaderexp/example-shaders/test_shader.wgsl
new file mode 100755
index 00000000..b8476ccd
--- /dev/null
+++ b/shaderexp/example-shaders/test_shader.wgsl
@@ -0,0 +1,40 @@
+struct UniformBufferObject {
+    resolution: vec2<f32>,
+    time: f32,
+}
+@group(0) @binding(0) var<uniform> ubo : UniformBufferObject;
+
+@stage(fragment) fn main(
+    @location(0) uv : vec2<f32>
+) -> @location(0) vec4<f32> {
+    let aspect = ubo.resolution.xy / ubo.resolution.y;
+    let translated_uv = (uv - vec2<f32>(0.5,0.5)) * 2.0 * aspect;
+    let freq:f32 = 5.0;
+    let speed: f32 = 5.0;
+    let h = (sin(freq * length(translated_uv) + speed * ubo.time) + 1.0) / 2.0;
+
+    let h_off = 20.0;
+    return vec4<f32>(hsl_to_rgb(h * (360.0 - h_off * 2.0) + h_off ,0.7,0.5),1.0);
+}
+
+// 0 ≤ H < 360, 0 ≤ S ≤ 1 and 0 ≤ L ≤ 1
+fn hsl_to_rgb(h:f32,s:f32,l:f32) -> vec3<f32> {
+    let tmp_h = h % 360.0;
+    let c = (1.0 - abs(2.0 * l - 1.0)) * s;
+    let x = c * (1.0 - abs((tmp_h / 60.0) % 2.0 - 1.0));
+    let m = l - c / 2.0;
+
+    let case_1 = vec3<f32>(c  ,x  ,0.0);
+    let case_2 = vec3<f32>(x  ,c  ,0.0);
+    let case_3 = vec3<f32>(0.0,c  ,x);
+    let case_4 = vec3<f32>(0.0,x  ,c);
+    let case_5 = vec3<f32>(x  ,0.0,c);
+    let case_6 = vec3<f32>(c  ,0.0,x);
+
+    return case_1 * f32(tmp_h < 60.0  && tmp_h >= 0.0) +
+           case_2 * f32(tmp_h < 120.0 && tmp_h >= 60.0) +
+           case_3 * f32(tmp_h < 180.0 && tmp_h >= 120.0) +
+           case_4 * f32(tmp_h < 240.0 && tmp_h >= 180.0) +
+           case_5 * f32(tmp_h < 300.0 && tmp_h >= 240.0) +
+           case_6 * f32(tmp_h < 360.0 && tmp_h >= 300.0) + vec3<f32>(m);
+}
diff --git a/shaderexp/frag.wgsl b/shaderexp/frag.wgsl
new file mode 100755
index 00000000..e13d4cbe
--- /dev/null
+++ b/shaderexp/frag.wgsl
@@ -0,0 +1,94 @@
+struct UniformBufferObject {
+    resolution: vec2<f32>,
+    time: f32,
+}
+@group(0) @binding(0) var<uniform> ubo : UniformBufferObject;
+
+fn getDist(p:vec3<f32>) -> f32{
+    let dist_from_center:f32 = 2.*sin(ubo.time * 3.);
+    let rotation_speed:f32 = 6.;
+    
+    let sphere1 = vec4<f32>(dist_from_center*cos(rotation_speed*ubo.time + 3.14159 * 0. * 2. / 3.),1.1, dist_from_center*sin(rotation_speed*ubo.time + 3.14159 + 3.14159 * 0. * 2. / 3.),1.);
+    let sphere2 = vec4<f32>(dist_from_center*cos(rotation_speed*ubo.time + 3.14159 * 1. * 2. / 3.),1.1, dist_from_center*sin(rotation_speed*ubo.time + 3.14159 + 3.14159 * 1. * 2. / 3.),1.);
+    let sphere3 = vec4<f32>(dist_from_center*cos(rotation_speed*ubo.time + 3.14159 * 2. * 2. / 3.),1.1, dist_from_center*sin(rotation_speed*ubo.time + 3.14159 + 3.14159 * 2. * 2. / 3.),1.);
+
+    let sphere1_dist:f32 = length(p - sphere1.xyz) - sphere1.w;
+    let sphere2_dist:f32 = length(p - sphere2.xyz) - sphere2.w;
+    let sphere3_dist:f32 = length(p - sphere3.xyz) - sphere3.w;
+    let plane_dist = p.y;
+    
+    return min(min(min(sphere1_dist,sphere2_dist),sphere3_dist),plane_dist);
+}
+
+fn rayMarch(ro:vec3<f32>, rd:vec3<f32>) -> f32{
+    let MAX_STEPS:i32 = 100;
+    let MAX_DIST:f32 = 100.0;
+    let SURF_DIST:f32 = 0.01;
+    var d:f32 = 0.0;
+    
+    var i: i32 = 0;
+    loop {
+      if(i >= MAX_STEPS){
+        break;
+      }
+
+      let p = ro + rd * d;
+      let ds = getDist(p);
+      d = d + ds;
+      if(d > MAX_DIST || ds <= SURF_DIST){
+        break;
+      } 
+
+      i = i + 1;
+    }
+    return d;
+}
+
+fn getNormal(p:vec3<f32>) -> vec3<f32>{
+    let d = getDist(p);
+    let e = vec2<f32>(0.1,0.0);
+    
+    // We can find the normal using the points around the hit point
+    let n = d - vec3<f32>(
+        getDist(p-e.xyy),
+        getDist(p-e.yxy),
+        getDist(p-e.yyx)
+      );
+        
+    return normalize(n);
+}
+
+fn getLight(p:vec3<f32>) -> f32{
+    let SURF_DIST:f32 = .01;
+
+    let light_pos = vec3<f32>(0.,5.,0.);
+    let l = normalize(light_pos - p) * 1.;
+    let n = getNormal(p);
+    
+    var dif = clamp(dot(n,l),.0,1.);
+    
+    let d = rayMarch(p + n * SURF_DIST * 2.,l);
+    if(d<length(light_pos - p)){
+        dif = dif * .1;
+    }
+    
+    return dif;
+}
+
+@stage(fragment) fn main(
+    @location(0) uv : vec2<f32>
+) -> @location(0) vec4<f32> {
+    let aspect = ubo.resolution / min(ubo.resolution.x,ubo.resolution.y);
+    let tmp_uv = (uv - vec2(0.5,0.5)) * aspect * 2.0;
+    var col = vec3<f32>(0.0);
+    
+    let r_origin = vec3<f32>(4.0,3.,.0);
+    let r_dir = normalize(vec3<f32>(-1.0,tmp_uv.y,tmp_uv.x));
+    let d = rayMarch(r_origin,r_dir);
+    col = vec3<f32>(d / 8.);
+    let p = r_origin + r_dir * d;
+    let diff = getLight(p);
+    
+    col = vec3<f32>(0.0, diff, 0.0);
+    return vec4<f32>(col,0.0);
+}
diff --git a/shaderexp/main.zig b/shaderexp/main.zig
new file mode 100755
index 00000000..ead03f37
--- /dev/null
+++ b/shaderexp/main.zig
@@ -0,0 +1,329 @@
+const std = @import("std");
+const mach = @import("mach");
+const gpu = @import("gpu");
+const glfw = @import("glfw");
+
+const App = mach.App(*FrameParams, .{});
+
+const Vertex = struct {
+    pos: @Vector(4, f32),
+    uv: @Vector(2, f32),
+};
+
+const vertices = [_]Vertex{
+    .{ .pos = .{ -1, -1, 0, 1 }, .uv = .{ 0, 0 } },
+    .{ .pos = .{ 1, -1, 0, 1 }, .uv = .{ 1, 0 } },
+    .{ .pos = .{ 1, 1, 0, 1 }, .uv = .{ 1, 1 } },
+    .{ .pos = .{ -1, 1, 0, 1 }, .uv = .{ 0, 1 } },
+};
+const indices = [_]u16{ 0, 1, 2, 2, 3, 0 };
+
+const UniformBufferObject = struct {
+    resolution: @Vector(2, f32),
+    time: f32,
+};
+
+var timer: std.time.Timer = undefined;
+
+pub fn main() !void {
+    timer = try std.time.Timer.start();
+    var gpa = std.heap.GeneralPurposeAllocator(.{}){};
+    var allocator = gpa.allocator();
+
+    const ctx = try allocator.create(FrameParams);
+    var app = try App.init(allocator, ctx, .{});
+
+    app.window.setKeyCallback(struct {
+        fn callback(window: glfw.Window, key: glfw.Key, scancode: i32, action: glfw.Action, mods: glfw.Mods) void {
+            _ = scancode;
+            _ = mods;
+            if (action == .press) {
+                switch (key) {
+                    .space => window.setShouldClose(true),
+                    else => {},
+                }
+            }
+        }
+    }.callback);
+    // On linux if we don't set a minimum size, you can squish the window to 0 pixels of width and height,
+    // this makes some strange effects when that happens, so it's better to leave a minimum size to avoid that,
+    // this doesn't prevent you from minimizing the window.
+    try app.window.setSizeLimits(.{ .width = 20, .height = 20 }, .{ .width = null, .height = null });
+
+    var fragment_file: std.fs.File = undefined;
+    var last_mtime: i128 = undefined;
+    if (std.fs.cwd().openFile("shaderexp/frag.wgsl", .{ .mode = .read_only })) |file| {
+        fragment_file = file;
+        if (file.stat()) |stat| {
+            last_mtime = stat.mtime;
+        } else |err| {
+            std.debug.print("Something went wrong when attempting to stat file: {}\n", .{err});
+            return;
+        }
+    } else |e| {
+        std.debug.print("Something went wrong when attempting to open file: {}\n", .{e});
+        return;
+    }
+    defer fragment_file.close();
+    var code = try fragment_file.readToEndAllocOptions(allocator, std.math.maxInt(u16), null, 1, 0);
+    defer allocator.free(code);
+
+    const queue = app.device.getQueue();
+
+    const vertex_buffer = app.device.createBuffer(&.{
+        .usage = .{ .vertex = true },
+        .size = @sizeOf(Vertex) * vertices.len,
+        .mapped_at_creation = true,
+    });
+    var vertex_mapped = vertex_buffer.getMappedRange(Vertex, 0, vertices.len);
+    std.mem.copy(Vertex, vertex_mapped, vertices[0..]);
+    vertex_buffer.unmap();
+    defer vertex_buffer.release();
+
+    const index_buffer = app.device.createBuffer(&.{
+        .usage = .{ .index = true },
+        .size = @sizeOf(u16) * indices.len,
+        .mapped_at_creation = true,
+    });
+    var index_mapped = index_buffer.getMappedRange(@TypeOf(indices[0]), 0, indices.len);
+    std.mem.copy(u16, index_mapped, indices[0..]);
+    index_buffer.unmap();
+    defer index_buffer.release();
+
+    // We need a bgl to bind the UniformBufferObject, but it is also needed for creating
+    // the RenderPipeline, so we pass it to recreatePipeline as a pointer
+    var bgl: gpu.BindGroupLayout = undefined;
+    const pipeline = recreatePipeline(&app, code, &bgl);
+
+    const uniform_buffer = app.device.createBuffer(&.{
+        .usage = .{ .copy_dst = true, .uniform = true },
+        .size = @sizeOf(UniformBufferObject),
+        .mapped_at_creation = false,
+    });
+    defer uniform_buffer.release();
+    const bind_group = app.device.createBindGroup(
+        &gpu.BindGroup.Descriptor{
+            .layout = bgl,
+            .entries = &.{
+                gpu.BindGroup.Entry.buffer(0, uniform_buffer, 0, @sizeOf(UniformBufferObject)),
+            },
+        },
+    );
+    defer bind_group.release();
+
+    ctx.* = FrameParams{
+        .pipeline = pipeline,
+        .queue = queue,
+        .vertex_buffer = vertex_buffer,
+        .index_buffer = index_buffer,
+        .uniform_buffer = uniform_buffer,
+        .bind_group = bind_group,
+
+        .fragment_shader_file = fragment_file,
+        .fragment_shader_code = code,
+        .last_mtime = last_mtime,
+    };
+
+    bgl.release();
+
+    try app.run(.{ .frame = frame });
+}
+
+const FrameParams = struct {
+    pipeline: gpu.RenderPipeline,
+    queue: gpu.Queue,
+    vertex_buffer: gpu.Buffer,
+    index_buffer: gpu.Buffer,
+    uniform_buffer: gpu.Buffer,
+    bind_group: gpu.BindGroup,
+
+    fragment_shader_file: std.fs.File,
+    fragment_shader_code: [:0]const u8,
+    last_mtime: i128,
+};
+
+fn frame(app: *App, params: *FrameParams) !void {
+    if (params.fragment_shader_file.stat()) |stat| {
+        if (params.last_mtime < stat.mtime) {
+            std.log.info("The fragment shader has been changed", .{});
+            params.last_mtime = stat.mtime;
+            params.fragment_shader_file.seekTo(0) catch unreachable;
+            params.fragment_shader_code = params.fragment_shader_file.readToEndAllocOptions(app.allocator, std.math.maxInt(u32), null, 1, 0) catch |err| {
+                std.log.err("Err: {}", .{err});
+                return;
+            };
+            params.pipeline = recreatePipeline(app, params.fragment_shader_code, null);
+        }
+    } else |err| {
+        std.log.err("Something went wrong when attempting to stat file: {}\n", .{err});
+    }
+
+    const back_buffer_view = app.swap_chain.?.getCurrentTextureView();
+    const color_attachment = gpu.RenderPassColorAttachment{
+        .view = back_buffer_view,
+        .resolve_target = null,
+        .clear_value = std.mem.zeroes(gpu.Color),
+        .load_op = .clear,
+        .store_op = .store,
+    };
+
+    const encoder = app.device.createCommandEncoder(null);
+    const render_pass_info = gpu.RenderPassEncoder.Descriptor{
+        .color_attachments = &.{color_attachment},
+        .depth_stencil_attachment = null,
+    };
+
+    const time = @intToFloat(f32, timer.read()) / @as(f32, std.time.ns_per_s);
+    const ubo = UniformBufferObject{
+        .resolution = .{ @intToFloat(f32, app.current_desc.width), @intToFloat(f32, app.current_desc.height) },
+        .time = time,
+    };
+    encoder.writeBuffer(params.uniform_buffer, 0, UniformBufferObject, &.{ubo});
+
+    const pass = encoder.beginRenderPass(&render_pass_info);
+    pass.setVertexBuffer(0, params.vertex_buffer, 0, @sizeOf(Vertex) * vertices.len);
+    pass.setIndexBuffer(params.index_buffer, .uint16, 0, @sizeOf(u16) * indices.len);
+    pass.setPipeline(params.pipeline);
+    pass.setBindGroup(0, params.bind_group, &.{0});
+    pass.drawIndexed(indices.len, 1, 0, 0, 0);
+    pass.end();
+    pass.release();
+
+    var command = encoder.finish(null);
+    encoder.release();
+
+    params.queue.submit(&.{command});
+    command.release();
+    app.swap_chain.?.present();
+    back_buffer_view.release();
+}
+
+fn recreatePipeline(app: *const App, fragment_shader_code: [:0]const u8, bgl: ?*gpu.BindGroupLayout) gpu.RenderPipeline {
+    const vs_module = app.device.createShaderModule(&.{
+        .label = "my vertex shader",
+        .code = .{ .wgsl = @embedFile("vert.wgsl") },
+    });
+    defer vs_module.release();
+    const vertex_attributes = [_]gpu.VertexAttribute{
+        .{ .format = .float32x4, .offset = @offsetOf(Vertex, "pos"), .shader_location = 0 },
+        .{ .format = .float32x2, .offset = @offsetOf(Vertex, "uv"), .shader_location = 1 },
+    };
+    const vertex_buffer_layout = gpu.VertexBufferLayout{
+        .array_stride = @sizeOf(Vertex),
+        .step_mode = .vertex,
+        .attribute_count = vertex_attributes.len,
+        .attributes = &vertex_attributes,
+    };
+
+    // Check wether the fragment shader code compiled successfully, if not
+    // print the validation layer error and show a black screen
+    app.device.pushErrorScope(.validation);
+    var fs_module = app.device.createShaderModule(&gpu.ShaderModule.Descriptor{
+        .label = "my fragment shader",
+        .code = .{ .wgsl = fragment_shader_code },
+    });
+    var error_occurred: bool = false;
+    // popErrorScope() returns always true, (unless maybe it fails to capture the error scope?)
+    _ = app.device.popErrorScope(&gpu.ErrorCallback.init(*bool, &error_occurred, struct {
+        fn callback(ctx: *bool, typ: gpu.ErrorType, message: [*:0]const u8) void {
+            if (typ != .noError) {
+                std.debug.print("🔴🔴🔴🔴:\n{s}\n", .{message});
+                ctx.* = true;
+            }
+        }
+    }.callback));
+    if (error_occurred) {
+        fs_module = app.device.createShaderModule(&gpu.ShaderModule.Descriptor{
+            .label = "my fragment shader",
+            .code = .{ .wgsl = @embedFile("black_screen_frag.wgsl") },
+        });
+    }
+    defer fs_module.release();
+
+    const blend = gpu.BlendState{
+        .color = .{
+            .operation = .add,
+            .src_factor = .one,
+            .dst_factor = .zero,
+        },
+        .alpha = .{
+            .operation = .add,
+            .src_factor = .one,
+            .dst_factor = .zero,
+        },
+    };
+    const color_target = gpu.ColorTargetState{
+        .format = app.swap_chain_format,
+        .blend = &blend,
+        .write_mask = gpu.ColorWriteMask.all,
+    };
+    const fragment = gpu.FragmentState{
+        .module = fs_module,
+        .entry_point = "main",
+        .targets = &.{color_target},
+        .constants = null,
+    };
+
+    const bgle = gpu.BindGroupLayout.Entry.buffer(0, .{ .fragment = true }, .uniform, true, 0);
+    // bgl is needed outside, for the creation of the uniform_buffer in main
+    const bgl_tmp = app.device.createBindGroupLayout(
+        &gpu.BindGroupLayout.Descriptor{
+            .entries = &.{bgle},
+        },
+    );
+    defer {
+        // In frame we don't need to use bgl, so we can release it inside this function, else we pass bgl
+        if (bgl == null) {
+            bgl_tmp.release();
+        } else {
+            bgl.?.* = bgl_tmp;
+        }
+    }
+
+    const bind_group_layouts = [_]gpu.BindGroupLayout{bgl_tmp};
+    const pipeline_layout = app.device.createPipelineLayout(&.{
+        .bind_group_layouts = &bind_group_layouts,
+    });
+    defer pipeline_layout.release();
+
+    const pipeline_descriptor = gpu.RenderPipeline.Descriptor{
+        .fragment = &fragment,
+        .layout = pipeline_layout,
+        .depth_stencil = null,
+        .vertex = .{
+            .module = vs_module,
+            .entry_point = "main",
+            .buffers = &.{vertex_buffer_layout},
+        },
+        .multisample = .{
+            .count = 1,
+            .mask = 0xFFFFFFFF,
+            .alpha_to_coverage_enabled = false,
+        },
+        .primitive = .{
+            .front_face = .ccw,
+            .cull_mode = .none,
+            .topology = .triangle_list,
+            .strip_index_format = .none,
+        },
+    };
+
+    // Create the render pipeline. Even if the shader compilation succeeded, this could fail if the
+    // shader is missing a `main` entrypoint.
+    app.device.pushErrorScope(.validation);
+    const pipeline = app.device.createRenderPipeline(&pipeline_descriptor);
+    // popErrorScope() returns always true, (unless maybe it fails to capture the error scope?)
+    _ = app.device.popErrorScope(&gpu.ErrorCallback.init(*bool, &error_occurred, struct {
+        fn callback(ctx: *bool, typ: gpu.ErrorType, message: [*:0]const u8) void {
+            if (typ != .noError) {
+                std.debug.print("🔴🔴🔴🔴:\n{s}\n", .{message});
+                ctx.* = true;
+            }
+        }
+    }.callback));
+    if (error_occurred) {
+        // Retry with black_screen_frag which we know will work.
+        return recreatePipeline(app, @embedFile("black_screen_frag.wgsl"), bgl);
+    }
+    return pipeline;
+}
diff --git a/shaderexp/vert.wgsl b/shaderexp/vert.wgsl
new file mode 100755
index 00000000..dbf0efbd
--- /dev/null
+++ b/shaderexp/vert.wgsl
@@ -0,0 +1,14 @@
+struct VertexOut {
+     @builtin(position) position_clip : vec4<f32>;
+     @location(0) frag_uv : vec2<f32>;
+}
+
+@stage(vertex) fn main(
+     @location(0) position : vec4<f32>,
+     @location(1) uv : vec2<f32>
+) -> VertexOut {
+     var output : VertexOut;
+     output.position_clip = position;
+     output.frag_uv = uv;
+     return output;
+}