diff --git a/Makefile b/Makefile
index 39cf94f0..91ad0735 100644
--- a/Makefile
+++ b/Makefile
@@ -6,6 +6,8 @@ src/collect_data:	src/collect_data.c Makefile
 src/preprocess:	src/preprocess.c Makefile
 	gcc -g -Wall -o src/preprocess src/preprocess.c
 
+src/make_error_tables:	src/make_error_tables.c Makefile
+	gcc -g -Wall -o src/make_error_tables src/make_error_tables.c
 
 newrom:	c64/kernal/OUT.BIN c64/basic/OUT.BIN
 	cat c64/basic/OUT.BIN c64/kernal/OUT.BIN  >newrom
diff --git a/src/make_error_tables.c b/src/make_error_tables.c
new file mode 100644
index 00000000..9e700065
--- /dev/null
+++ b/src/make_error_tables.c
@@ -0,0 +1,162 @@
+#include <stdio.h>
+#include <strings.h>
+#include <string.h>
+#include <stdlib.h>
+#include <unistd.h>
+
+// XXX - Find out which of these are standard error messages, and thus not subject
+// to any copyright claims, versus any that are specific to Commodore, and thus
+// could be copyright.
+char *error_list[]={
+  "TOO MANY FILES",
+  "FILE OPEN",
+  "FILE NOT OPEN",
+  "FILE NOT FOUND",
+  "DEVICE NOT PRESENT",
+  "NOT INPUT FILE",
+  "NOT OUTPUT FILE",
+  "MISSING FILENAME",
+  "ILLEGAL DEVICE NUMBER",
+  "NEXT WITHOUT FOR",
+  "SYNTAX",
+  "RETURN WITHOUT GOSUB",
+  "OUT OF DATA",
+  "ILLEGAL QUANTITY",
+  "OVERFLOW",
+  "OUT OF MEMORY",
+  "UNDEF\'D STATEMENT",
+  "BAD SUBSCRIPT",
+  "REDIM\'D ARRAY",
+  "DIVISION BY ZERO",
+  "ILLEGAL DIRECT",
+  "TYPE MISMATCH",
+  "STRING TOO LONG",
+  "FILE DATA",
+  "FORMULA TOO COMPLEX",
+  "CAN\'T CONTINUE",
+  "UNDEF\'D FUNCTION",
+  "VERIFY",
+  "LOAD",
+  NULL};
+
+// Token $FF is end of message, so we can have only 255 unique words
+#define MAX_WORDS 255
+char *words[MAX_WORDS];
+int word_count=0;
+
+#define MAX_LEN 1024
+unsigned char message_tokens[MAX_LEN];
+int token_count=0;
+
+struct char_freq {
+  char c;
+  int count;
+};
+
+int cmpfreq(const void *a,const void *b)
+{
+  const struct char_freq *aa=a,*bb=b;
+  if (aa->count<bb->count) return 1;
+  if (aa->count>bb->count) return -1;
+  return 0;
+}
+
+struct char_freq char_freqs[31];
+int char_max=0;
+
+int find_word(char *w)
+{
+  int i;
+  for(i=0;i<word_count;i++) {
+    if (!strcmp(w,words[i])) {
+      //      fprintf(stderr,"Word '%s' is repeated.\n",w);
+      return i;
+    }
+  }
+  words[word_count++]=strdup(w);
+  return word_count-1;
+}
+
+void process_word(char *w)
+{
+  int num=find_word(w);
+  message_tokens[token_count++]=num;
+  return;
+}
+
+void end_of_message(void)
+{
+  message_tokens[token_count++]=0xFF;
+}
+
+int main(void)
+{
+  int raw_size=0;
+
+  // Get letter frequencies
+  for(int i=0;i<26;i++) char_freqs[i].count=0;
+  for(int i=0;error_list[i];i++) {
+    for(int j=0;error_list[i][j];j++)
+      if (error_list[i][j]!=' ') {
+	int char_num=0;
+	for(char_num=0;char_num<char_max;char_num++)
+	  if (error_list[i][j]==char_freqs[char_num].c) break;
+	if (char_num>(14+16)) {
+	  fprintf(stderr,"Too many unique characters in message list.  Max is 30\n");
+	  exit(-1);
+	}
+	if (char_num>=char_max) char_max=char_num+1;
+	char_freqs[char_num].c=error_list[i][j];
+	char_freqs[char_num].count++;
+      }
+  }
+  qsort(char_freqs,char_max,sizeof(struct char_freq),cmpfreq);
+  fprintf(stderr,"Letter frequencies:\n");
+  for(int i=0;i<char_max;i++) fprintf(stderr,"'%c' : %d\n",char_freqs[i].c,char_freqs[i].count);
+
+  // Get 14 most frequent out to use as 4-bit tokens 1 - 14.
+  // Token 0 = end of word.
+  // Token 15 indicates opposite nybl identifies which of the other 16 possible values is used
+  // for less-frequent letters.
+  // Thus common letters take 4 bits, and uncommon ones take 8.
+  
+  
+  for(int i=0;error_list[i];i++) raw_size+=strlen(error_list[i]);
+  fprintf(stderr,"Error list takes %d bytes raw.\n",raw_size);
+
+  for(int i=0;error_list[i];i++)
+    {
+      char word[1024];
+      int offset=0;
+      for(int j=0;error_list[i][j];j++) {
+	if (error_list[i][j]==' ') {
+	  // Found word boundary
+	  int w=0;
+	  for(int k=offset;k<j;k++) word[w++]=error_list[i][k];
+	  word[w++]=0;
+	  offset=j+1;
+	  //	  fprintf(stderr,"Found word '%s'\n",word);
+	  process_word(word);
+	}
+      }
+      {
+	// Get last word on line
+	int w=0;
+	for(int k=offset;k<error_list[i][k];k++) word[w++]=error_list[i][k];
+	word[w++]=0;
+	//	fprintf(stderr,"Found word '%s'\n",word);
+	process_word(word);
+	end_of_message();
+      }
+    }	
+
+  fprintf(stderr,"%d token bytes used to encode all messages.\n",token_count);
+
+  int word_bytes=0;
+  for(int i=0;i<word_count;i++) word_bytes+=strlen(words[i]);
+  fprintf(stderr,"Plus %d bytes for the words\n",word_bytes);
+
+  fprintf(stderr,"Space saving = %d bytes\n",raw_size-word_bytes-token_count);
+	  
+  
+}