Experiment 96
0018, 0027, 0062 and 0065 simulated one-stage recipes and asked about one fingerprint at a time. Here the recipes are the two-stage constructions of experiments 80 to 88 (a K3-style move before or after a repeating key, the self-feeding autokey, and 0087's column-plus-row key), and the question is the joint one: how often does a recipe make the width-7 doublet column and the width-21 repeats in the same 97 letters?
What the method does with the 24 known letters.
K4the carved letters The 97 letters cut into the copper at the CIA, shown at the top of every experiment page. has two odd patterns: when written in rows of 7, its doubled lettersthe doublets Places where a carved letter is immediately followed by the same letter. stack in one column, and when written in rows of 21, letter pairs repeat straight down more than chance allows. A good guess at the method should make both patterns at once. We build thousands of fake coppers with each two-step recipe tried in experiments 80 to 88, a grida grid Writing the message into rows of some width and reading it out in another order, column by column for instance, which shuffles the letters without changing them. move and the tablethe table The right-hand half of the sculpture, 26 rows of the alphabet each shifted one place further, starting from the word KRYPTOS., and count how often each recipe makes both patterns together.
Words in the margin are defined on the glossary page.
The carved letters go in at the top; the reading comes out at the bottom.
K4 has two oddities: five of its six doubled letters fall in one column of a 7-wide grid, and in rows of 21 eleven letter pairs repeat straight down. Either alone turns up by chance now and then.
Twenty-nine recipes, moves and keys in both orders, each enciphered 20,000 stretches of English. A recipe that made K4 should produce both marks together far more often than chance.
No recipe produces both more than twice in 20,000. The column-plus-row key at width 7 makes the doublet column only at the chance rate, so 0087's premise that it explains the column does not hold.
This experiment produced no candidate: the search ruled everything out before a full text was built. Its product is the bounded negative in section 7.
Redo the method on a few letters with pencil and paper.
The same steps run by the code on the real letters.
K4's two marks
doublets (place of first letter) and place mod 7:
BB at 19-20: column 4
QQ at 26-27: column 4
SS at 33-34: column 4
SS at 43-44: column 0
ZZ at 47-48: column 4
TT at 68-69: column 4
letter pairs 21 apart that repeat: 11
rows of 21:
OBKRUOXOGHULBSOLIFBBW
FLRVQQPRNGKSSOTWTQSJQ
SSEKZZWATJKLUDIAWINFB
NYPVTTMZFPKWGDKZXTJCD
IGKUHUAUEKCAR
Family, key space, the constraint under test, and the clues it reads.
Type 2 of directions.md, open item 'a joint simulation that makes both fingerprints'. 0056: the width-7 doublet column (5 of 6 doublets in one column) and the width-21 vertical repeats (11 repeated letter pairs 21 apart) survive the look-elsewhere correction (corrected p 0.005 and 0.010); 0062 found ciphertext autokey at lag 21 raises the repeats, 0065 found ciphertext autokey at lag 1 with a period-7 key raises the column, and 0087 found that key = a[j mod 7] + b[j div 7] lines doublets up by construction. Recipes (each 20,000 samples of 97 letters from Carter vol. 1 and K1-K3, random keys): periodic key 7, 14 or 21 with a width-7 columnar, a width-7 quarter turn or a K3-rule move, before or after; ciphertext autokey lag 1 or lag 21 with a period-7 key, with a width-7 move before; the column-plus-row key at width 7 and 21, with and without a move before; and calibrations (random letters, shuffled K4, plain period-7 key). Marks as in 0065 (column: at least 4 doublets and all but one in one column at width 7) and 0056 (repeats: at least K4's 11 at width 21).
Written before the run; the result decides it.
No recipe makes both marks in more than 1% of samples. The column-plus-row key at width 7 makes the doublet column often (over 10%) but the width-21 repeats only at the chance rate; ciphertext autokey at lag 21 makes the repeats but not the column. Zero candidates.
The harness's record: candidates, gates, loss, and the negative result this experiment leaves behind.
Rates column / repeats / both: width-7 columnar, then ciphertext autokey lag 1 + period-7 key: 0.0086 / 0.0019 / 0.0001; K3-rule move, then period-14 key: 0.0023 / 0.0054 / 0.0001; width-7 quarter turn, then period-14 key: 0.0022 / 0.0029 / 0.0001; K3-rule move, then column-plus-row key, width 7: 0.0021 / 0.0022 / 0.0001; period-14 key, then width-7 quarter turn: 0.0019 / 0.0043 / 0.0001; calibration: period-7 key alone: 0.0019 / 0.0147 / 0.0001; width-7 quarter turn, then period-21 key: 0.0017 / 0.0040 / 0.0001; width-7 columnar, then ciphertext autokey lag 21 + period-7 key: 0.0013 / 0.0029 / 0.0001; period-21 key, then width-7 quarter turn: 0.0011 / 0.0033 / 0.0001; ciphertext autokey lag 1 + period-7 key: 0.0140 / 0.0024 / 0.0000; period-7 key, then width-7 quarter turn: 0.0035 / 0.0137 / 0.0000; column-plus-row key, width 7: 0.0026 / 0.0014 / 0.0000.
What was searched, as the harness recorded it.
How this experiment sits among the others is on the lineage page; its family's open hypotheses are on the families page.