“A column shuffle combined with the scrambled text as its own key, in either order”
Experiment 62 found that a cipher keyed by its own scrambled output, 21 letters back, is the only recipe tried that makes the width-21 repeats. On its own that cipher is ruled out by the known letters (experiment 2). Here it is combined with a column shuffle, before or after, which changes where the known letters fall, and tested exactly.
In a ciphertextthe carved letters The 97 letters cut into the copper at the CIA, shown at the top of every experiment page.autokeyan autokey cipher A cipher that uses the message itself, or the scrambled text itself, as the secret word, each letter enciphered with the letter a fixed distance before it., each letter of the messagethe message The 97 letters of English that Jim Sanborn wrote and enciphered. is hidden using the hidden letter that came some fixed distance earlier, so the cipher feeds on its own output. Experiment 62 showed that this recipe, at a distance of 21, leaves the same 21-letter mark the copper has. On its own it cannot produce the copper's known letters, but a shuffle of columns before or after it would move them. Here we try every shuffle in our list, every distance and every way of using 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 check whether the 24 known lettersthe known letters Sanborn has confirmed that positions 22 to 34 of the message read EASTNORTHEAST and positions 64 to 74 read BERLINCLOCK. come out where they should.
Words in the margin are defined on the glossary page.
1.1
The method at work
The carved letters go in at the top; the reading comes out at the bottom.
the idea
Experiment 62 found that ciphertext autokey at distance 21 makes the copper's width-21 repeats. On its own it cannot produce the known letters, so this asks whether a column shuffle before or after it can.
the start
In a ciphertext autokey each letter is keyed by the carved letter a fixed distance before it, so the first 21 letters have an unknown key. Known letters that land there cannot be checked.
the bound
3,107 shuffles, distances 1 to 95, all 24 ways of using the table, both orders: 0 fits, best 9 of 24.
2
Best candidate against the known letters
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.
3
Walkthrough, by hand
Redo the method on a few letters with pencil and paper.
Pick lag 21 and the sculpture's alphabet with Vigenere.
Undo the autokey on the copper: the letter at place 22 is the carved letter at 22 minus the carved letter at place 1 on the table, F minus O.
That gives one letter of the shuffled message.
If the shuffle were rows of 7 read down the columns, the known E at place 22 would have been moved to place 4, inside the first 21 letters where the autokey's starter is unknown, so it says nothing.
Check every other known letter at its new place past 21.
One mismatch kills this shuffle, lag and table.
the worksheet: the first steps by hand, on squared paper
4
Worked example, generated
The same steps run by the code on the real letters.
lag 21, kry/kry/kry/vig, autokey undone on the copper (first 21 unknown):
?????????????????????ACRUZIOVSZOIXZZSIBXBWQIDTOOQSEPKEIOEDNUGVHGVMVOOUMBGKSMKEMRIWWYVEWZDLIVZWCDH
5
Hypothesis, for a cryptanalyst
Family, key space, the constraint under test, and the clues it reads.
Ciphertext autokey (key = ciphertext letter L back, read in the key alphabet) at every lag 1-96 under the 24 conventions, composed with a transposition. Order A: transposition of the plaintext first, then autokey; the autokey decryption of K4 is then the transposed plaintext, computable without any key except the L-letter primer, so each crib must reappear at its route image (images inside the primer are free). Order B: autokey first, then transposition; undoing the route and then the autokey gives the plaintext directly, so the cribs are checked in place. Routes: 0007's set (plain columnar widths 2-96 both reading directions and both senses, keyed columnar from each vocabulary word, double plain columnar 2-50) plus keyed columnar from K1-K3 and C13/C14 words. Bounds: exact for these routes, lags and conventions; motivated by the width-21 repeats (0056, 0062).
6
Prediction
Written before the run; the result decides it.
No route, lag and convention reproduces all 24 cribs in either order; the best matches stay at a few cribs. Zero candidates.
7
Method
For each lag and convention, undo the autokey on K4 (order A) and record the text.
For each route, check that every crib whose image lies past the primer matches.
For order B, undo each route on K4 first, then the autokey, and check the cribs in place.
Keep the best counts.
Decrypt and submit any full match.
8
Result and bounds
The harness's record: candidates, gates, loss, and the negative result this experiment leaves behind.
3107 routes x 95 lags x 24 conventions x 2 orders (14,167,920 checks): 0 full matches. Best order A: 8 cribs past the primer (('A', 'double columnar w=12 then w=49', 16, 'kry/kry/kry/vig')); best order B: 9/24 (('B', 'double columnar w=34 then w=23', 13, 'kry/kry/kry/varbeau')).
What was searched, as the harness recorded it.
family
transposition composed with ciphertext autokey, either order
routes
3,107
lags
1-95
conventions
24
checked
14,167,920
full matches
0
best order A
hits past primer
8
setting
A, double columnar w=12 then w=49, 16, kry/kry/kry/vig
best order B
hits
9
setting
B, double columnar w=34 then w=23, 13, kry/kry/kry/varbeau
How this experiment sits among the others is on the lineage page; its family's open hypotheses are on the families page.
Rendered from the record on 2026-10-18 12:26 UTC; updated after every experiment.