diff --git a/README.md b/README.md
index 9426a876..6df00813 100644
--- a/README.md
+++ b/README.md
@@ -2,7 +2,7 @@
 
 [![Build Status](https://travis-ci.org/FluxML/Flux.jl.svg?branch=master)](https://travis-ci.org/FluxML/Flux.jl) [![](https://img.shields.io/badge/docs-stable-blue.svg)](https://fluxml.github.io/Flux.jl/stable/) [![Join the chat at https://gitter.im/FluxML](https://badges.gitter.im/FluxML/Lobby.svg)](https://gitter.im/FluxML/Lobby) [Slack](https://discourse.julialang.org/t/announcing-a-julia-slack/4866)
 
-Flux is a library for machine learning, implemented in Julia. Flux is high-level yet extremely lightweight, providing only simple abstractions on top of Julia's native GPU support and automatic differentiation.
+Flux is an unusually elegant machine learning library. It provides lightweight abstractions on top of Julia's native GPU and AD support, while remaining fully hackable (right down to the [GPU kernels](https://github.com/FluxML/CuArrays.jl)).
 
 Define a simple model using any Julia code: