SIDE QUEST 01 / CRYPTANALYSIS / RESEARCH LOG
A public-facing field notebook for my attempt to understand the mechanism behind the 97-character K4 passage. The interesting part is not a guessed plaintext; it is the progressive reduction of a huge hypothesis space into precise, falsifiable families.
Twenty-four plaintext positions are anchored by the known EAST, NORTHEAST, BERLIN and CLOCK clues. I use those anchors as constraints, then ask whether a precisely defined cipher family can satisfy them and still recover coherent language elsewhere. A null result is useful only when the tested family and search completeness are explicit.
Within the crib windows, ciphertext equals plaintext at positions 32 and 73: S→S and K→K. For Quagmire III, a conjugated non-zero shift has no fixed points, so these positions force zero keystream states. That creates a strong algebraic filter before any language scoring.
An exact feasibility screen of the recorded tied keyword/composite Quagmire III family tested 864,788 systems and returned no survivors. That closes that bounded construction—not every possible QIII mechanism.
The largest conceptual update is a coincidence invariant that survives arbitrary alphabet substitutions. The once-central single-layer period-8 architecture is statistically excluded at about P ≈ 2×10⁻⁵. This is stronger than failing to find a good key: it screens the entire fixed-state bijective family under the stated English model.
A Round-5 review found an incomplete QIII enumerator: fixing the first key state to zero was not a valid gauge for that tie mode. Earlier progressive-key “eliminations” were withdrawn and rerun. Planted controls and completeness tests are part of the result, not optional polish.
This visual summarizes the current crib-consistency × coincidence-invariant map from the research record. It is a screening result, not a plaintext solution.
A bijective substitution can rename letters, but it cannot change which letters are equal. Group positions by a proposed state schedule and compare the within-state equality pattern with natural English. This tests arbitrary alphabets without guessing either alphabet.
Period 8 was the architecture around which several earlier rounds were organized because it is the smallest weak periodic chart compatible with the cribs. The invariant later showed that this attractive minimal model is statistically inconsistent with a single-layer fixed-state bijection of natural English. That negative result made many expensive period-8 searches retrospectively unnecessary.
Follow branches as they are tested, narrowed, corrected, eliminated or kept alive.
This interface summarizes recorded experiments from the Kryptos research repository. It does not claim a K4 solution; statuses apply only to the precisely tested model families.
Change the period to regroup all 97 positions by index modulo n. The two self-encrypting crib positions, 32 and 73, are especially informative for Quagmire III.
Treating W as null changes the schedule. n=19 survives the coincidence screen and remains a particularly interesting decidable lead.
Before optimizing alphabets, test properties that survive arbitrary substitutions. Coincidence structure can reject whole periodic families without searching millions of keys.
A model can satisfy all four crib blocks and still decrypt the other 73 letters into noise. Exact constraint solving answers “can this family fit?”; controlled language scoring asks whether it explains the rest.
At 97 characters, adding free alphabets, masks or long periods can make a model flexible enough to fit almost anything. The research therefore tracks residual degrees of freedom and prioritizes families that the available evidence can actually distinguish.
Gromark, two-clock schedules, physical/tableau-derived streams and null hypotheses are useful leads, but a suggestive score is not a solution. “Disfavoured,” “open,” and “eliminated” are intentionally different labels.
The strongest remaining work is not another unconstrained brute-force sweep. It is targeted: the W-null QIII n=19 Vigenère rerun, one-slip QIII cells, unresolved two-clock cells, QI/QII progressive and period-24 cases, selected key-tied known streams, and exact tied autokeys. Families beyond the information content of K4 are labeled OPEN-K5 rather than forced into a conclusion. The research repository remains the detailed computational record; this site is the visual explanation layer.