diff --git a/README.md b/README.md
index da61be6..842d6a9 100644
--- a/README.md
+++ b/README.md
@@ -1,13 +1,18 @@
 # Paychex payroll importer for Beancount
 
-The aim of this tool is to automate the monthly task of transcribing payroll CSV
-information for Conservancy's 8 (currently) employees into ~ 300 lines of fairly
-intricate Beancount bookkeeping data.
+This tool automates the monthly task of transcribing a payroll CSV from service
+provider Paychex into around 300 lines of fairly intricate Beancount bookkeeping
+data.
 
 
 ## Usage
 
-See a demo with:
+The run, the program requires a Java runtime (tested with OpenJDK 17 and 21).
+On Debian 12 (Bookworm):
+
+    sudo apt install openjdk-17-jre
+
+Run a demo with two example employees, Jack and Jill Citizen:
 
     java -jar import-x.x.x-standalone.jar --demo
 
@@ -16,8 +21,8 @@ Provide your own payroll data with:
     java -jar import-x.x.x-standalone.jar --csv resources/example-paychex-pay-item-details.csv --total-fees 206.50
 
 In the above, various values such as the date, time period covered and
-receipt/invoice values will be replace with "TODO" placeholders. You can
-alternatively provide any/all of these explicitly:
+receipt/invoice values show "TODO" placeholders that you are expected to fill in
+later. If you prefer, you can provide any/all of these explicitly:
 
     java -jar import-x.x.x-standalone.jar --csv resources/example-paychex-pay-item-details.csv --date 2023-12-29 --period 'December 2023' --total-fees 206.50 --pay-receipt-no rt:19462/674660 --pay-invoice-no rt:19403/675431 --fees-receipt-no rt:19459/675387 --fees-invoice-no rt:19459/674887 --retirement-receipt-no rt:19403/676724 --retirement-invoice-no rt:19403/675431
 
@@ -37,6 +42,21 @@ Then run Beancount with:
     bean-report [your file] balances
 
 
+## Building
+
+To build you need a Java SDK (tested with OpenJDK 17 and 21) and Clojure. On
+Debian 12 (Bookworm):
+
+    sudo apt install openjdk-17-jdk clojure
+
+To build, run:
+
+    bin/build
+
+This will output a JAR file like `target/import-x.x.x-standalone.jar`, where the
+version number is based on the Git revision.
+
+
 ## Development
 
 Run tests with:
@@ -51,10 +71,15 @@ The project is set up for development in Emacs and CIDER-mode. Open a source
 file and run `cider-jack-in`.
 
 
-## Build
+## Why Clojure?
 
-To build, run:
-
-    bin/build
-
-This will output a JAR file like `target/import-x.x.x-standalone.jar`.
+Clojure is very well-suited to this kind of data
+transformation/manipulation. Even without fully understanding the implications
+of all the accounting/tax concepts I was able to parse the previous
+human-generated Beancount entries into a data structure and work backwards to
+produce results that matched exactly, without worrying about exact text
+formatting. Libraries like deep-diff2 and the editor-integrated REPL made it
+very efficient to compare the data structures and build the program - certainly
+much more efficient than developing the same program in Python. It's only just
+over 500 lines of code and we've had zero bugs as of writing (one year since it
+was built).
diff --git a/TODO.md b/TODO.md
index 7bf01f7..cd2148d 100644
--- a/TODO.md
+++ b/TODO.md
@@ -1,5 +1,5 @@
 # To-do
 
-* Validation/error handling on CLI arguments - required options like csv, date, period
+* Add validation/error handling of required CLI arguments such as csv, date and period
 * Potentially move employee-name->entity-tag into config
 * Potentially move cat->payroll-type into config
diff --git a/bin/build b/bin/build
index a16f8ef..21d40ce 100755
--- a/bin/build
+++ b/bin/build
@@ -1,2 +1,4 @@
 #!/usr/bin/env sh
+
+# Build an uberjar (a Java jar package including the program and all dependencies)
 clojure -T:build uber
diff --git a/bin/dev b/bin/dev
index 9d984aa..cd7dc20 100755
--- a/bin/dev
+++ b/bin/dev
@@ -1,2 +1,4 @@
 #!/usr/bin/env sh
+
+# Run the program without building it
 clojure -M -m core "$@"
diff --git a/bin/loc b/bin/loc
new file mode 100755
index 0000000..b6df469
--- /dev/null
+++ b/bin/loc
@@ -0,0 +1,8 @@
+#!/usr/bin/env sh
+
+# Count the lines of code, excluding tests and examples
+git ls-files -- \
+    ':!:LICENSE.txt' \
+    ':!:test/*' \
+    ':!:resources/example-*.csv' \
+    | xargs cloc "$@"
diff --git a/bin/test b/bin/test
index af4849e..6eeaf1d 100755
--- a/bin/test
+++ b/bin/test
@@ -1,2 +1,4 @@
 #!/usr/bin/env sh
+
+# Run the test suite
 clojure -M:test "$@"
diff --git a/manifest.scm b/manifest.scm
new file mode 100644
index 0000000..e258fa9
--- /dev/null
+++ b/manifest.scm
@@ -0,0 +1,12 @@
+;; Guix manifest
+;;
+;; If you're running GNU Guix, this file describes the dependencies necessary to
+;; develop/build the program. Use it with `guix shell --manifest=manifest.scm`.
+(specifications->manifest
+ (list
+  ;; No issues running this OpenJDK 21 program on Debian Stable (OpenJDK 17).
+  "openjdk@21"
+  ;; Works fine with clojure-tools from Guix.
+  "clojure-tools"
+  "beancount"
+  ))