diff --git a/README.md b/README.md
index 7ec7140..c3527ee 100644
--- a/README.md
+++ b/README.md
@@ -18,19 +18,28 @@ join in--send patches, help add stuff, etc.
 
 ## The method
 
+Documenting this for other texts to be converted in future...
+
+First we need to extract the text documents from the disks and turn them into
+something we can use on a modern system:
+
 1. The DOS 3.3 disks were dumped using cppo
-2. Apply the following transformations to each document file:
+2. Apply `scripts/extract_piewriter.py` to each document file which did the
+   following transformations:
    * For characters 0xa0-0xfe, strip the high bit to get pure ASCII
    * Convert 0x0d and 0x8d (return) characters ti 0x0a (newline)
    * Escape all else in C-style
-3. Remove NUL at end of .txt files
-4. .pp dot command is paragraph break, replace with blank line.
-5. Remove trailing whitespace
-6. Normalize case and spacing of dot commands (lowercase here)
+3. Remove NUL at end of .txt files and renamed the assembly source to .s
+4. Remove trailing whitespace
+6. Normalize dot commands (lowercase, spacing) for easier mechanical parsing.
+7. Remove the obvious dot commands (.pp is a paragraph break, .sp creates
+   vertical space, .br seems to be a line break, .bp a page break) and attempt
+   to remove or interpret others as seems appropriate
 
-
-This has probably broken the .s files a bit, and I haven't bothered to decompile
-the five byte HELLO ...  ;)
+This process has probably broken the .s files and there were some files that
+don't appear to have actually been part of the text (or maybe they were edits
+and revisions?), and there was bitrot in the files suggesting the disks the
+source documents were stored on were losing their integrity.
 
 [dons-disks]: http://www.6502lane.net/2015/03/12/don-worths-beneath-apple-dos-original-text-files/
 [archive.org]: https://archive.org/
diff --git a/scripts/extract_piewriter.py b/scripts/extract_piewriter.py
new file mode 100755
index 0000000..31d88df
--- /dev/null
+++ b/scripts/extract_piewriter.py
@@ -0,0 +1,40 @@
+#! /usr/bin/env python3
+
+"""extract_piewriter.py <filename> [<filename>...]
+
+Extracts PIEWriter documents extracted as raw "#064000" (binary blob) files
+from Apple DOS 3.3 disks.  Performs the following conversions:
+
+ - Strips high bits from printable ASCII characters that have it set.
+ - Converts Mac-style CR-delimited lines to UNIX-style LF-delimited.
+ - Replaces any other character with its C-style escaped hex representation
+   (e.g., NUL is replaced with \\x00)
+
+The output is rough, but its enough to check it in to a git repository and
+begin cleaning up now properly text files.
+"""
+
+import sys
+
+if len(sys.argv) == 1:
+    print(sys.modules[__name__].__doc__)
+    sys.exit(1)
+
+for arg in sys.argv[1:]:
+    with open(arg, 'rb') as f:
+        infile = f.read()
+
+    outfile = bytearray()
+
+    for val in infile:
+        if 0xa0 <= val < 0xff:
+            outfile.append(val & 0x7f)
+        elif val in (0x0d, 0x8d, 0x8a):
+            outfile.append(0x0a)
+        else:
+            outfile.extend('\\x{:02x}'.format(val).encode('ASCII'))
+
+    outname = ''.join((arg, '.txt'))
+    print('Saving', outname)
+    with open(outname, 'wb') as f:
+        f.write(outfile)