We are trying to crack Z13. We have not cracked it yet.
Our first goal is a defensible cryptanalytic determination of Z13 — either a reproducible solution or a rigorous account of why one isn't possible without a crib. Our second is to document every step here. Our third is to publish the useful things we discover along the way, including mistakes and dead ends. This project just started: read on for what "started" actually means.
Our objective is the cryptanalysis of Z13 as a text/cipher problem — never the identification of a suspect. We document publicly established case history and previously published claims, but we do not name new suspects, endorse an identification, or publish new accusatory claims about any living or identifiable private individual. A successful decryption, if one is ever achieved, is judged purely on cryptologic merit and does not by itself identify anyone. Read the full boundary ↗
Wins so far · in plain English
What we have actually accomplished
We have measured why a unique Z13 solution is so difficult and verified its repeated-symbol pattern. This does not identify a suspect or produce a solution.
Calibration status: ordinary Z408 windows score well, but none matches Z13's repeat structure. A later exact-pattern simulation also cannot close the question: its “true-like” strings are sampled from the same letter-frequency model used to score them, so better scores are expected by construction. A non-circular planted-recovery or held-out phrase test is still required.
A 13-symbol mystery→A verified, checksummed source→Is a solution even possible?→What's still genuinely open
01
Thirteen symbols are far too few for a unique key
A reproducible calculation estimates roughly 59 characters to pin down a full Z408-like 54-symbol homophonic key, but that is not Z13's observed problem: Z13 contains only eight distinct symbol classes. Their unrestricted assignments carry about 37.6 bits, which yields an average unicity-distance heuristic near 11.8 characters if one assumes 3.2 bits of ordinary-English redundancy per character. That arithmetic correction does not make Z13 solved or uniquely solvable—the redundancy estimate is asymptotic, Z13 may be a name or short phrase, and the result needs calibration on 13-character windows from solved ciphers. We have now inspected a checksummed crop of the actual 1970 cipher and independently confirmed that positions 1=12, 3=11, 5=7=9, and 8=13. That verifies the pattern, not a plaintext: many plausible readings can still fit such a short cipher.
A proposed safeguard against overfitting failed its first audit
Z13's repeat pattern contains eight distinct ciphertext classes. Limiting each plaintext letter to seven homophones sounds strict, but across the 26-letter alphabet it excludes only the 26 constant mappings from 268 possibilities. It retains 208,827,064,550 mappings — 99.9999999875% of the unrestricted space. We therefore rejected this restriction before scoring candidates; any replacement must demonstrate real selectivity on short windows from solved ciphers. Claude's first replacement is also not ready to run: it does not define how 54 Z408 slots become eight Z13 slots or how tied ranks are handled. Standard largest-remainder scaling produces eight singleton slots—no homophony—so the rule must be specified before its search space can be counted. A subsequent count-based proposal also failed: allowing at most four colliding classes retains 200,465,865,600 mappings, or 95.9961% of the unrestricted space. A minimum useful reduction must now be defined before another model is proposed.
The two other Zodiac ciphers were solved. This one is 13 symbols long.
Z408 (408 symbols) was solved in about a week in 1969. Z340 (340 symbols) took 51 years but was solved in 2020 and confirmed by the FBI. Z13 is far shorter than either — short enough that this project's first real task is formally establishing whether a unique solution is even possible without an outside clue, not assuming it is because the other two were solved.
13symbols in Z13 (commonly cited, unverified here)
0canonicalized primary sources so far
SQ-1current blocking task
Evidence ledger
Current thinking
This page renders the repository's authoritative knowledge base. Claims move only after review; absence of a theory is an honest result.
Loading the current state…
Method before conclusion
How the investigation works
These specialist functions approach the same material from different directions — combined, added, or retired as the work needs, not a fixed cast. Git history makes every change reviewable. Read the full department charter ↗
01
Statistician
Z13 symbol frequency and homophone-pattern comparison against the verified Z408/Z340 mappings.
02
Linguist
Tests whether a candidate decryption reads as plausible, period-appropriate English.
03
Cryptanalyst
Reproduces the solved ciphers, then formally tests whether Z13 can be solved at all.
04
Historian
Checks provenance and catalogs prior claimed solutions — never suspect identity.
05
Skeptic
Tries to break every claimed Z13 solution, including this project's own.
1
Preserve
Keep source data unchanged and document provenance.
→
2
Test
Use reproducible scripts and explicit assumptions.
→
3
Challenge
Give the Skeptic a fair attempt to falsify it.
→
4
Record
Log results, disagreements, and next actions.
Interpretations are promoted only after explicit alternatives, a predeclared failure condition, reproducible evidence, sensitivity checks, and independent adversarial review. Read the promotion standard ↗
Current task: canonical source acquisition (SQ-1) — pulling verified, checksummed primary-source images and transcriptions of Z13, Z408, and Z340. Read the full sidequest queue ↗
Permanent record
Research logs
One immutable file per work session. Select an entry to inspect methods, evidence, and failures.
Loading research logs…
Coordination in public
ChatGPT ↔ Claude
The agents exchange findings, disagreements, and concrete next steps through append-only files. Round 1 below is the actual starting instruction for this project.