diff --git a/.gitignore b/.gitignore
index 65d47b80..1e9ab3f4 100644
--- a/.gitignore
+++ b/.gitignore
@@ -5,3 +5,4 @@
 .swp
 dist
 ollama
+ggml-metal.metal
diff --git a/.gitmodules b/.gitmodules
index e2142b3e..cd623ca6 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -1,3 +1,4 @@
 [submodule "llm/llama.cpp/ggml"]
 	path = llm/llama.cpp/ggml
 	url = https://github.com/ggerganov/llama.cpp.git
+	ignore = dirty
diff --git a/README.md b/README.md
index 4796f042..eb41fdf4 100644
--- a/README.md
+++ b/README.md
@@ -127,19 +127,26 @@ Ollama bundles model weights, configuration, and data into a single package, def
 
 ## Building
 
-You will also need a C/C++ compiler such as GCC for MacOS and Linux or Mingw-w64 GCC for Windows.
+Install `cmake`:
 
 ```
+brew install cmake
+```
+
+Then generate dependencies and build:
+
+```
+go generate ./...
 go build .
 ```
 
-To run it start the server:
+Next, start the server:
 
 ```
-./ollama serve &
+./ollama serve
 ```
 
-Finally, run a model!
+Finally, run a model in another shell:
 
 ```
 ./ollama run llama2