diff --git a/README.md b/README.md
index 12b8305..9fcd27b 100644
--- a/README.md
+++ b/README.md
@@ -1,23 +1,73 @@
 loot-condition-interpreter
 ==========================
 
-A library for parsing and evaluating LOOT's metadata condition strings. This
-library improves upon LOOT's existing implementation in the following ways:
+[![AppVeyor Build Status](https://ci.appveyor.com/api/projects/status/github/loot/loot-condition-interpreter?branch=master&svg=true)](https://ci.appveyor.com/project/loot/loot-condition-interpreter)
+[![Travis Build Status](https://travis-ci.org/loot/loot-condition-interpreter.svg?branch=master)](https://travis-ci.org/loot/loot-condition-interpreter)
+[![dependency status](https://deps.rs/repo/github/Ortham/libloadorder/status.svg)](https://deps.rs/repo/github/loot/loot-condition-interpreter)
 
-- Condition expressions are parsed into an intermediate representation instead
-  of being stringly typed, which allows parsing and evaluation to be separated.
-- State is uncoupled by necessity, and Rusts's concurrency guarantees mean it
-  can be accessed more efficiently.
-- Results are cached with more granularity, per function instead of per
-  expression, improving performance when expressions are not entirely different.
-- Result caching is guided by benchmarks, so results aren't cached
-  unnecessarily.
-- Reading executable versions doesn't involve calling out to the shell and
-  piping several commands together when on Linux.
+A library for parsing and evaluating LOOT's metadata condition strings. It
+provides:
 
-The code is also not as much of a mess, it's got benchmarks, and probably better
-test coverage.
+- Condition string parsing without evaluation, for checking syntax.
+- Evaluation of parsed condition strings.
+- Efficient and safe concurrent condition evaluation thanks to Rust's safety 
+  guarantees.
+- Caching of individual function evaluation results and calculated CRCs.
+- Executable version parsing without any external runtime dependencies.
+- Lots of tests, and benchmarks.
+- A C FFI library that wraps the Rust library.
 
-The tests need the [testing-plugins](https://github.com/WrinklyNinja/testing-plugins)
+## Build
+
+Make sure you have [Rust](https://www.rust-lang.org/) installed.
+
+To build the Rust and C FFI libraries, run:
+
+```
+cargo build --release --package loot-condition-interpreter-ffi
+```
+
+To generate a C++ header file as part of the build, instead run:
+
+```
+cargo build --release --package loot-condition-interpreter-ffi --all-features
+```
+
+This will create a file at `ffi/include/loot_condition_interpreter.hpp`.
+However, this involves building [cbindgen](https://github.com/eqrion/cbindgen)
+as a dependency, which is relatively slow. If you have cbindgen installed
+separately, it's faster to generate the headers using its CLI:
+
+```
+cbindgen ffi/ -l c++ -o ffi/include/loot_condition_interpreter.hpp
+```
+
+## Tests & Benchmarks
+
+The tests and benchmarks need the [testing-plugins](https://github.com/WrinklyNinja/testing-plugins)
 and the [LOOT API v0.13.8](https://github.com/loot/loot-api/releases/tag/0.13.8)
-Windows archives to be extracted and present in the repo root.
+Windows archives to be extracted and present in the repo root. See the AppVeyor
+and Travis CI configurations for examples on what should be extracted where.
+
+To run the Rust tests:
+
+```
+cargo test --all
+```
+
+To run the benchmarks:
+
+```
+cargo bench
+```
+
+There are also C++ tests for the FFI library, they require a C++ toolchain and
+[CMake](https://cmake.org/) to be installed. To run the C++ tests:
+
+```
+mkdir ffi/build
+cd ffi/build
+cmake ..
+cmake --build .
+ctest
+```