diff --git a/assets/shaders/pbr.vert.spv b/assets/shaders/pbr.vert.spv
index 3330c72..5f2ca88 100644
Binary files a/assets/shaders/pbr.vert.spv and b/assets/shaders/pbr.vert.spv differ
diff --git a/src/gears/game.d b/src/gears/game.d
index 50e8994..f019bd4 100644
--- a/src/gears/game.d
+++ b/src/gears/game.d
@@ -68,9 +68,10 @@ InitGame(PlatformWindow* window)
 		input_rate_stride: Vertex.sizeof,
 		vertex_attributes: [
 			{ binding: 0, location: 0, format: FMT.RGBA_F32, offset: 0 },
-			{ binding: 0, location: 1, format: FMT.RGBA_F32, offset: Vertex.n.offsetof },
-			{ binding: 0, location: 2, format: FMT.RG_F32, offset: Vertex.uv.offsetof },
-			{ binding: 0, location: 3, format: FMT.RGBA_UNORM, offset: Vertex.col.offsetof },
+			{ binding: 0, location: 1, format: FMT.RGBA_F32, offset: Vertex.tangent.offsetof },
+			{ binding: 0, location: 2, format: FMT.RGB_F32, offset: Vertex.pos.offsetof },
+			{ binding: 0, location: 3, format: FMT.RGB_F32, offset: Vertex.normal.offsetof },
+			{ binding: 0, location: 4, format: FMT.RG_F32, offset: Vertex.uv.offsetof },
 		],
 	};
 
diff --git a/src/gears/main.d b/src/gears/main.d
index d434556..c786d77 100644
--- a/src/gears/main.d
+++ b/src/gears/main.d
@@ -7,6 +7,7 @@ import game;
 import util;
 import core.simd;
 import math;
+import core.stdc.string : memcpy;
 
 // TODO: 
 // 1. Remove bindless (informed to be perf death)
@@ -24,7 +25,12 @@ void main(string[] argv)
 	while (true)
 	{
 		HandleEvents(&window);
-		if (window.close) break;
+		if (window.close) 
+		{
+			Logf("quitting");
+			break;
+		}
+
 
 		Cycle(&g);
 	}
diff --git a/src/gears/renderer.d b/src/gears/renderer.d
index a8a30a7..59e3230 100644
--- a/src/gears/renderer.d
+++ b/src/gears/renderer.d
@@ -118,20 +118,6 @@ struct Renderer
 	PushConst push_const;
 }
 
-extern(C) struct Material
-{
-	Vec4 ambient;
-	Vec4 albedo;
-	Vec4 specular;
-	u32 albedo_texture;
-	u32 ambient_texture;
-	u32 specular_texture;
-	b32 albedo_has_texture;
-	b32 ambient_has_texture;
-	b32 specular_has_texture;
-	f32 shininess = 0.0;
-}
-
 struct Extent
 {
 	u32 x;
@@ -148,10 +134,11 @@ struct UIVertex
 
 struct Vertex
 {
-	Vec4 pos;
-	Vec4 n;
+	Vec4 color;
+	Vec4 tangent;
+	Vec3 pos;
+	Vec3 normal;
 	Vec2 uv;
-	u32 col;
 }
 
 struct MeshPart
@@ -214,18 +201,6 @@ GetExtent(Renderer* rd)
 	return Extent(rd.vk.swapchain_extent.width, rd.vk.swapchain_extent.height);
 }
 
-pragma(inline): void
-CopyVertex(Vec4* dst, m3dv_t* src)
-{
-	asm
-	{
-		mov R8, src;
-		mov R9, dst;
-		movups XMM0, src.x.offsetof[R8];
-		movups dst.x.offsetof[R9], XMM0;
-	}
-}
-
 pragma(inline): void
 BeginFrame(Renderer* rd)
 {
@@ -465,34 +440,37 @@ LoadModel(Renderer* rd, string name)
 	foreach(i; 0 .. m3d.numface)
 	{
 		u32 vi = (i * 3);
+		u32 i0 = vi+0;
+		u32 i1 = vi+1;
+		u32 i2 = vi+2;
 
-		CopyVertex(&vertices[vi+0].pos, &m3d.vertex[m3d.face[i].vertex[0]]);
-		CopyVertex(&vertices[vi+1].pos, &m3d.vertex[m3d.face[i].vertex[1]]);
-		CopyVertex(&vertices[vi+2].pos, &m3d.vertex[m3d.face[i].vertex[2]]);
+		CopyVertex(&vertices[i0].pos, &m3d.vertex[m3d.face[i].vertex[0]]);
+		CopyVertex(&vertices[i1].pos, &m3d.vertex[m3d.face[i].vertex[1]]);
+		CopyVertex(&vertices[i2].pos, &m3d.vertex[m3d.face[i].vertex[2]]);
 
-		CopyVertex(&vertices[vi+0].n, &m3d.vertex[m3d.face[i].vertex[0]]);
-		CopyVertex(&vertices[vi+1].n, &m3d.vertex[m3d.face[i].vertex[1]]);
-		CopyVertex(&vertices[vi+2].n, &m3d.vertex[m3d.face[i].vertex[2]]);
+		CopyVertex(&vertices[i0].normal, &m3d.vertex[m3d.face[i].normal[0]]);
+		CopyVertex(&vertices[i1].normal, &m3d.vertex[m3d.face[i].normal[1]]);
+		CopyVertex(&vertices[i2].normal, &m3d.vertex[m3d.face[i].normal[2]]);
 
-		vertices[vi+0].col = m3d.vertex[m3d.face[i].vertex[0]].color;
-		vertices[vi+1].col = m3d.vertex[m3d.face[i].vertex[1]].color;
-		vertices[vi+2].col = m3d.vertex[m3d.face[i].vertex[2]].color;
+		ConvertColor(&vertices[i0].color, m3d.vertex[m3d.face[i].vertex[0]].color);
+		ConvertColor(&vertices[i1].color, m3d.vertex[m3d.face[i].vertex[1]].color);
+		ConvertColor(&vertices[i2].color, m3d.vertex[m3d.face[i].vertex[2]].color);
 
 		if (m3d.numtmap)
 		{
-			vertices[vi+0].uv.x = m3d.tmap[m3d.face[i].texcoord[0]].u;
-			vertices[vi+0].uv.y = m3d.tmap[m3d.face[i].texcoord[0]].v;
+			vertices[i0].uv.x = m3d.tmap[m3d.face[i].texcoord[0]].u;
+			vertices[i0].uv.y = m3d.tmap[m3d.face[i].texcoord[0]].v;
 
-			vertices[vi+1].uv.x = m3d.tmap[m3d.face[i].texcoord[1]].u;
-			vertices[vi+1].uv.y = m3d.tmap[m3d.face[i].texcoord[1]].v;
+			vertices[i1].uv.x = m3d.tmap[m3d.face[i].texcoord[1]].u;
+			vertices[i1].uv.y = m3d.tmap[m3d.face[i].texcoord[1]].v;
 
-			vertices[vi+2].uv.x = m3d.tmap[m3d.face[i].texcoord[2]].u;
-			vertices[vi+2].uv.y = m3d.tmap[m3d.face[i].texcoord[2]].v;
+			vertices[i2].uv.x = m3d.tmap[m3d.face[i].texcoord[2]].u;
+			vertices[i2].uv.y = m3d.tmap[m3d.face[i].texcoord[2]].v;
 		}
 
-		indices[vi+0] = vi+0;
-		indices[vi+1] = vi+1;
-		indices[vi+2] = vi+2;
+		indices[i0] = i0;
+		indices[i1] = i1;
+		indices[i2] = i2;
 	}
 
 	assert(vertices.length > 0 && indices.length > 0, "Error loading model");
@@ -511,6 +489,14 @@ LoadModel(Renderer* rd, string name)
 	return model;
 }
 
+pragma(inline): void
+CopyVertex(Vec3* dst, m3dv_t* src)
+{
+	dst.x = src.x;
+	dst.y = src.y;
+	dst.z = src.z;
+}
+
 void
 Destroy(Renderer* rd)
 {
diff --git a/src/gears/vulkan.d b/src/gears/vulkan.d
index b3a1566..b935e6c 100644
--- a/src/gears/vulkan.d
+++ b/src/gears/vulkan.d
@@ -283,6 +283,11 @@ Init(PlatformWindow* window, u64 permanent_mem, u64 frame_mem)
 		value: vk,
 	};
 
+	if (!success)
+	{
+		Logf("error initializing vulkan");
+	}
+
 	return result;
 }
 
diff --git a/src/packer/packer.d b/src/packer/packer.d
index 0408b7b..b2d893f 100644
--- a/src/packer/packer.d
+++ b/src/packer/packer.d
@@ -172,7 +172,7 @@ TestFile()
 	}
 }
 
-u32
+static u32
 MagicValue(string str)
 {
 	assert(str.length == 4, "Magic value must 4 characters");
@@ -192,3 +192,60 @@ InitHeader(u64 asset_count)
 	return header;
 }
 
+static ModelHeader
+InitModelHeader()
+{
+	ModelHeader header = {
+		magic: MagicValue("stgm"),
+		model_version: MODEL_VERSION,
+	};
+
+	return header;
+}
+
+ModelData
+ConvertModel(u8[] data)
+{
+	assert(data.size > 4, "Model data too small for a magic number");
+
+	u32 magic = *cast(u32*)(data.ptr);
+
+	ModelData model;
+	switch (magic)
+	{
+		case MagicValue("3DMO"):
+			{
+				model = ConvertM3D(data);
+			} break;
+		case MagicValue("glTF"):
+			{
+				model = ConvertGLTF(data);
+			} break;
+		default:
+			{
+				assert(false, "Unsupported model type");
+			} break;
+	}
+
+	return model;
+}
+
+ModelData
+ConvertGLTF(u8[] data)
+{
+	ModelData model;
+
+	assert(false, "Not yet supported");
+
+	return model;
+}
+
+ModelData
+ConvertM3D(u8[] data)
+{
+	ModelData model;
+
+	return model;
+}
+
+
diff --git a/src/shaders/pbr.vert.glsl b/src/shaders/pbr.vert.glsl
index 83b793f..1e37a61 100644
--- a/src/shaders/pbr.vert.glsl
+++ b/src/shaders/pbr.vert.glsl
@@ -5,10 +5,11 @@
 
 #include "structures.layout"
 
-layout (location = 0) in vec4 in_pos;
-layout (location = 1) in vec4 in_normal;
-layout (location = 2) in vec2 in_uv;
-layout (location = 3) in vec4 in_col;
+layout (location = 0) in vec4 in_col;
+layout (location = 1) in vec4 in_tangent;
+layout (location = 2) in vec3 in_pos;
+layout (location = 3) in vec3 in_normal;
+layout (location = 4) in vec2 in_uv;
 
 layout (location = 0) out vec4 out_col;
 layout (location = 1) out vec2 out_uv;
@@ -22,7 +23,7 @@ mat4 y_matrix = mat4(
 
 void main()
 {
-	gl_Position = G.projection_matrix * G.view_matrix * PC.model_matrix * vec4(in_pos.rgb, 1.0);
+	gl_Position = G.projection_matrix * G.view_matrix * PC.model_matrix * vec4(in_pos, 1.0);
 
 	vec4 col = Materials[nonuniformEXT(PC.mat_id)].diffuse;
 
diff --git a/src/shared/aliases.d b/src/shared/aliases.d
index 13d3428..f3e7345 100644
--- a/src/shared/aliases.d
+++ b/src/shared/aliases.d
@@ -29,6 +29,8 @@ alias b32 = uint;
 alias intptr = intptr_t;
 alias uintptr = uintptr_t;
 
+alias usize = size_t;
+
 alias Vec2 = Vector!(f32, 2);
 alias Vec3 = Vector!(f32, 3);
 alias Vec4 = Vector!(f32, 4);
diff --git a/src/shared/assets.d b/src/shared/assets.d
index 5ae66b3..574aab6 100644
--- a/src/shared/assets.d
+++ b/src/shared/assets.d
@@ -14,6 +14,7 @@ AssetInfo[] Asset_Info;
 u8[][] Asset_Data;
 
 const u32 FILE_VERSION = 2;
+const u32 MODEL_VERSION = 1;
 
 enum AssetType : u32
 {
@@ -33,6 +34,64 @@ struct FileHeader
 	u64 asset_info_offset;
 }
 
+struct ModelHeader
+{
+	u32 magic;
+	u32 model_version;
+	u64 vertex_count;
+	u64 vertex_offset;
+	u64 index_count;
+	u64 index_offset;
+	u64 material_count;
+	u64 material_offset;
+	u64 texture_count;
+	u64 texture_offset;
+}
+
+struct Vertex
+{
+	Vec4 color;
+	Vec4 tangent;
+	Vec3 pos;
+	Vec3 normal;
+	Vec2 uv;
+}
+
+struct ModelData
+{
+	Vertex[] 			vertices;
+	u32[]		 			indices;
+	Material[] 		materials;
+	TextureInfo[] textures;
+}
+
+struct Material
+{
+	Vec4 ambient;
+	Vec4 albedo;
+	Vec4 specular;
+	u32 albedo_texture;
+	u32 ambient_texture;
+	u32 specular_texture;
+	b32 albedo_has_texture;
+	b32 ambient_has_texture;
+	b32 specular_has_texture;
+	f32 shininess = 0.0;
+}
+
+struct TextureInfo
+{
+	string name;
+	u32 	 id;
+}
+
+struct TextureHeader
+{
+	u64 str_length;
+	u64 str_offset;
+	u32 texture_id;
+}
+
 struct ModelMeta
 {
 	u64 index_count;
diff --git a/src/shared/util.d b/src/shared/util.d
index 4db439f..ecbb50a 100644
--- a/src/shared/util.d
+++ b/src/shared/util.d
@@ -558,3 +558,69 @@ GenerateLoop(string format_string, int N)() nothrow pure @safe
 	return result;
 }
 
+void
+MemCpy(void* dst_p, void* src_p, u64 length)
+{
+	u8* dst = cast(u8*)dst_p;
+	u8* src = cast(u8*)src_p;
+
+	u64 remaining = length;
+	if (remaining >= 64)
+	{
+		for(u64 i = 0; i < length; i += 64)
+		{
+			asm
+			{
+				mov R8, src;
+				mov R9, dst;
+
+				add R8, i;
+				movdqu XMM0, [R8+00];
+				movdqu XMM1, [R8+16];
+				movdqu XMM2, [R8+32];
+				movdqu XMM3, [R8+48];
+		
+				add R9, i;
+				movdqu [R9+00], XMM0;
+				movups [R9+16], XMM1;
+				movups [R9+32], XMM2;
+				movups [R9+48], XMM3;
+
+				sub remaining, 64;
+			}
+		}
+	}
+
+	if (remaining >= 32)
+	{
+		for(u64 i = remaining; i < length; i += 32)
+		{
+			asm
+			{
+				mov R8, src;
+				mov R9, dst;
+
+				add R8, i;
+				movdqu XMM0, [R8+00];
+				movdqu XMM1, [R8+16];
+
+				add R9, i;
+				movdqu [R9+00], XMM0;
+				movdqu [R9+16], XMM1;
+
+				sub remaining, 32;
+			}
+		}
+	}
+
+	for(u64 i = remaining; i < length; i += 1)
+	{
+		dst[i] = src[i];
+	}
+}
+
+void
+MemSet(void* dst, char ch, usize length)
+{
+
+}