diff --git a/README.md b/README.md
index fc285f3..9b022e8 100644
--- a/README.md
+++ b/README.md
@@ -28,11 +28,12 @@ Learning/Testing, results are inaccurate.
 The following scripts are in the repo:
 
 * `wut` --- Feed it an observation ID and it returns if it is a "good", "bad", or "failed" observation.
+* `wut-dl-sort` --- Populate `data/` dir with waterfalls from `download/`.
+* `wut-ml` --- Main machine learning Python script using Tensorflow and Keras.
 * `wut-obs` --- Download the JSON for an observation ID.
+* `wut-review-staging` --- Review all images in `data/staging`.
 * `wut-water` --- Download waterfall for an observation ID to `download/[ID]`.
 * `wut-water-range` --- Download waterfalls for a range of observation IDs to `download/[ID]`.
-* `wut-ml` --- Main machine learning Python script using Tensorflow and Keras.
-* `wut-review-staging` --- Review all images in `data/staging`.
 
 
 # Installation
@@ -99,8 +100,8 @@ The following steps need to be performed:
    These get put in the `downloads/[ID]/` directories.
 
 1. Organize downloaded waterfalls into categories (e.g. "good", "bad", "failed").
-   Note: this needs a script written.
-   Put them into their respective directories under:
+   Use `wut-dl-sort` script.
+   The script them into their respective directories under:
 	* `data/train/good/`
 	* `data/train/bad/`
 	* `data/train/failed/`
diff --git a/wut-dl-sort b/wut-dl-sort
new file mode 100755
index 0000000..02abfcc
--- /dev/null
+++ b/wut-dl-sort
@@ -0,0 +1,64 @@
+#!/bin/bash
+# wut-dl-sort
+#
+# Populates the data/ directory from the download/dir.
+#
+# XXX  This script removes directories in data/  !!! XXX
+#
+# Usage:
+# wut-dl-sort [Minimum Observation ID] [Maximum Observation ID]
+# Example:
+# wut-dl-sort 1292441 1470517
+#
+# * Takes the files in the download/ dir.
+# * Looks at the JSON files to see if it is :good", "bad", or "failed".
+# * Hard link it in the appropriate data/ directory.
+# * File is randomly copied to either data/train or data/val directory.
+#
+# Possible vetted_status: bad, failed, good, null, unknown.
+
+OBSIDMIN="$1"
+OBSIDMAX="$2"
+OBSID=$OBSIDMIN
+
+# First make sure we have the range all downloaded:
+echo "Downloading Waterfalls"
+./wut-water-range $OBSIDMIN $OBSIDMAX
+
+# XXX remove data/train and data/val directories XXX
+echo "Removing data/ subdirectories"
+rm -rf data/train data/val
+# Create new empty dirs
+mkdir -p data/train/good data/train/bad data/train/failed
+mkdir -p data/val/good   data/val/bad   data/val/failed
+
+# Then parse each file and link appropriately
+echo "Parsing download/ directory for observation IDs $OBSIDMIN to $OBSIDMAX"
+cd download/ || exit
+
+while [ $OBSID -lt $OBSIDMAX ]
+        do echo "ID: $OBSID "
+	cd $OBSID
+	VET=`cat $OBSID.json | jq --compact-output '.[0] | {vetted_status}' | cut -f 2 -d ":" | sed -e 's/}//g' -e 's/"//g'`
+	RAND_DIR=`echo $((0 + RANDOM % 2))`
+	if [ $RAND_DIR = 1 ] ; then
+		CLASS_DIR="train"
+	else
+		CLASS_DIR="val"
+	fi
+	case "$VET" in
+		bad) ln waterfall_$OBSID_*.png ../../data/$CLASS_DIR/$VET/
+		;;
+		good) ln waterfall_$OBSID_*.png ../../data/$CLASS_DIR/$VET/
+		;;
+		failed) ln waterfall_$OBSID_*.png ../../data/$CLASS_DIR/$VET/
+		;;
+		null) echo "null, not copying"
+		;;
+		unknown) echo "unknown, not copying"
+		;;
+	esac
+
+        let OBSID=$OBSID+1
+	cd ..
+done