diff --git a/README.md b/README.md
index d0fb170..5c42ce7 100644
--- a/README.md
+++ b/README.md
@@ -1,18 +1,18 @@
-
+
Get given random crystal structures predicted by ML materials discovery projects, and vote on whether you think it should be synthesizable (or, at least, whether it is worth trying!)
Currently uses [cifplayer](https://github.com/tilde-lab/cifplayer) and data from the
[OPTIMADE API](https://optimade.org) for the recent [Google DeepMind
paper](https://github.com/google-deepmind/materials_discovery).
-The physical properties prediction is done dynamically by [MPDS](https://mpds.io/ml).
+The physical properties prediction is done dynamically by [MPDS ML server](https://mpds.io/ml).
## Build
-The `cifplayer` is powered by `$mol`. Note that, unlike many other frontend frameworks, `$mol` provides the same single environment for all its projects, under the standard namespace scheme. That is, all your `$mol`-based code lives inside the same directory `$MOL_HOME`. So if you don't have `$MOL_HOME` yet, please create it and navigate there:
+The `cifplayer` is powered by `$mol` framework. Note that, unlike many other frontend frameworks, `$mol` provides the same single environment for all its projects, under the standard namespace scheme. That is, all your `$mol`-based code lives inside the same directory `$MOL_HOME`. So if you don't have `$MOL_HOME` yet, please create it and navigate there:
```bash
mkdir $MOL_HOME && cd $MOL_HOME