Update:

Using longer parts of the cipher somehow led to unsatisfying results..long (FCCP-) strings show up with tens of thousand results containing e.g. two cleartext words..cross-checking them with other strings quickly ends up in a memory error.

However, two strings are quite interesting..starting from the last symbol of line 2 (length 13 symbols) as well as the 3rd symbol of line 9 (length 14 symbols). Although both strings do not contain many homophones of the opposite string, the string from line 9 contains 12 (out of 14) symbols identical to line 2 or deriving from repeating bigrams, thus 86% of line 9 can be illustrated by the settings of line 2.

This is great news because now – so far without getting any memory error – both lines can be cross-checked against each other for valid cleartext.

First run & results (ever) is ongoing..although still a long way to, there now is a promising cracking procedure possible.

My personal opinion is still somehow divided:
On one hand, billions of homephone settings are eliminated, with two partial cleartext strings showing up as a result and the potential to continue the cross-checking procedure with other parts of the cipher (leading to a potential completion of the cleartext).
On the other hand uncommon bigrams might repeat as symbols and therefore influence the initial setting of line 2 for the solving process negatively.

In any case, it’s worth a try. Most I like we don’t talk about figuring out only one single symbol anymore. And with a bit luck, in some months..

QT

Another update..

Regarding the following strings

one may find that most of the homophones are either

– the ‘plus’ symbol
– part of the ‘IoFBc’ (the only two overlapping, repeating trigrams)
– part of a repeating bigram
– or present in both strings

This allows us to set up a run with Python. Assumptions are the 340 being a homophone substitution (no matter what sequences have been used) as well as that the repeating bigrams/trigrams are frequent ones (no statistical outlyers such as YJX).

Doing so, we can now set up a range of bigrams and trigrams, latter one with the requirement of being able to overlap with one letter. We also may set plus symbol (e.g. ‘L’ or ‘S’). By this, the 3.6e+90 possible letter combinations can be reduced dramatically to approximately 2.9e+12 or 2,900B combinations.

We then complete the strings regarding all symbols which do not meet the criteria mentioned above. Eg. the , which may represent any of the A-Z letters, is now represented by such a letter. Our result is a fictitiously created cleartext phrase (FCCP). This, of course, is not only a single phrase but a vector of multiple phrase variations that may occurr under the given settings.

In our case we have two FCCP strings (or set of strings..it’s all about set theory, after all). Those two FCCP sets can now be cross-checked for cleartext words. Only those constellations in which both strings contain an English word is of interest for us. For practical reasons it may be required to search for words of a certain length, e.g. 4 or 5 letters.

Now all computation – we still talk about 2,900B combinations for each type of ‘plus’ symbol/letter – can be done. Step-by-step, on a normal pc. And the runs are completely independent from any homophone sequences, too.

If there are more results, e.g. 200 million of such constellations, those can – again – be cross-checked with a different, third string. However, before doing so, a total of 138 letters or 40.5% of the cipher are already ‘cleared’ (meaning that if you had the correct constellation in front of you, it most likely would be solvable like a SF Chronicle cross-word riddle). With less than 200,000 of such constellations, manual proceeding with Excel would be possible.

QT

Here we go with some of the first results (still not fully working but on its way..):

QT

Update..

It was possible to increase from two strings to three strings

Thus, the program has to find three different cleartext parts in different cipher areas. For one 5-gram (IoFBc) only, the computation takes approximately 6 hours. The first two quintgrams analysed were ‘THERE’ and ‘ERERE’ in combination with ‘L’ as the + symbol.

About 90,000 values showed up of which most of them contain similar or equal cleartext words (~20 to 30 different word combinations). The following is an example of results:

As one can see, the program was able to find three cleartext words in three different strings of the cipher. The words are not colliding but match the setting of homophones in the 340. Therefore, it could already be part of Z’s real cleartext. However they still could have occurred accidentially (hill top) as the rest of the strings is rather no cleartext.

Please note that for the first string, besides IoFBc and the + symbol, a total of seven letters have been added (mainly from repeating bigrams). For the second string, only two new letters have been added. Also, for the third string only two additional letters have been added. This means that it is not possible to simply write multiple different words into the strings but are the strings all strictly related to each other (you may want to try..). IMO, it’s the first time ever that we can see such potential cleartext in this constellation.

Computation will proceed with the next 5-grams.

QT