Experiment 118
0002 tested ciphertext autokey with the first key letters free, so known places within one lag of the start gave no check. On the copper K4 is not a separate text: it runs on from K3 on the same plate. Here the key for place i is the carved letter lag places earlier in the whole carved stream (K1, K2, K3 and the ? as carved), so every known place is a check, for every lag 1-200; and the geometric version, the letter carved directly above on the previous plate line, and two lines above.
What the method does with the 24 known letters.
In a self-feeding key, each letter is locked with a letter that was already carved some distance before. Experiment 2 tried this starting fresh at K4the carved letters The 97 letters cut into the copper at the CIA, shown at the top of every experiment page.. But on the sculpture K4 just continues the lines of K3the earlier messages The three solved parts of the sculpture., so the first letters of K4 could take their key from K3's letters. We try every distance, and also the letter carved straight above on the line before, against all 24 known letters.
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.
In a self-feeding key each letter is locked with a letter carved some distance before. Experiment 2 started this fresh at K4, but on the sculpture K4 continues K3's lines, so its first letters could take their key from K3.
For each distance from 1 to 200, or the letter straight above on the plate, carved letter minus key letter must give the known letter at all 24 places.
At best 5 of 24 by distance (scrambled coppers 5-6) and 3 of 24 for the letter above (coppers 2-5): no setting comes near.
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.
the carved stream before K4 has 772 characters; lag 31 (one carved line), sculpture's alphabet, Vigenere
place known carved key letter (31 back) carved - key hit
22 E F T B no
23 A L T G no
24 S R R K no
25 T V T M no
26 N Q V X no
64 B N S G no
Family, key space, the constraint under test, and the clues it reads.
C7 ('it will come back to the first part'; earlier text holds instructions), C3(a) (the carved plate as the grid), C11a (NSA 1992: autokey). Model: c_i = E(p_i, x_{i-lag}) where x is the carved stream K1+K2+K3+?+K4 as on the copper (with or without the ? characters, which can carry no key value and are skipped in one variant), lag 1-200; 12 conventions (Vigenere, Beaufort, variant; plain and cipher alphabets standard or KRYPTOS; key letters on the cipher alphabet). Geometric variant: the key for a K4 letter is the letter in the same column of the line above (or two above) on the lower-left plate, columns counted from the left and from the right. Every known place gives a check: 24 per setting. Ten shuffled coppers (K4 shuffled, the earlier text fixed).
Written before the run; the result decides it.
No lag or geometric reading matches more than 6 of 24 known letters; same as the coppers. Zero candidates.
The harness's record: candidates, gates, loss, and the negative result this experiment leaves behind.
Lags 1-200 x 12 rules x 2 streams: K4 best 5/24, controls [5, 5, 5, 5, 6, 5, 5, 6, 5, 6]; full 0. Letter above on the plate: K4 best 3/24, controls [4, 4, 4, 4, 5, 4, 4, 2, 4, 2].
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.