Experiment 8

“K3's own matrix run through all the possible shifts, then the tableau”

Route set replaced. Instead of columnar grids, the routes are K3's transposition rule with every shift, plus K3-style double rotations of a grid. Survival test unchanged.

revertedstatus
experiment 6parents
2 mintime, all phases
0candidates
0/24known letters
1

The experiment method

What the method does with the 24 known letters.

The third messagethe earlier messages The three solved parts of the sculpture. was scrambled by a fixed rule: take every 192nd letter, wrapping around. The artist once said he was surprised nobody had taken that same rule and tried all its settings. So we do: every step size and every starting point for the 97 letters, and also the grid-turning trick the code-breakers described for the third message. For each setting we work out where 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. would have gone and ask whether a short secret word on the sculpture's tablethe table The right-hand half of the sculpture, 26 rows of the alphabet each shifted one place further, starting from the word KRYPTOS. could then produce the carved lettersthe carved letters The 97 letters cut into the copper at the CIA, shown at the top of every experiment page. there. A setting that allows a short word is the signal.

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.

OBKRUOXOGHULBSOLIFBBWEASTNORTHEASTOTWTQSJQSSEKZZWATJKLUDIAWINFBBERLINCLOCKWGDKZXTJCDIGKUHUAUEKCARIN: THE COPPER, ALL 97 LETTERS, THE 24 KNOWN PLACES IN ACCENTstart at 5, step 7: message place p is carved at 5 + 7(p - 1), wrapping past 97EASTNORTHEASTBERLINCLOCKplace 22 comes from 5 + 7 x 21 = 152, which wraps to 55the shortest secret word each rule allows, over 504,600 rule and table settings223040506070809422: the shortest any rule allows (4 settings)OUT: NO READING; NO PUZZLE WORD FITS ANY RULE'S FORCED LETTERSEASTNORTHEASTBERLINCLOCK
the rule

K3 was solved by reading its letters with a fixed start and step. Here every start and step is tried before the table: drawn for start 5, step 7, where the E at place 22 comes from carved place 55.

the forced letters

Each known letter and the carved letter it lands on force one key letter. Two places whose distance is a multiple of the key length share a slot, so their letters must agree.

the histogram

Over 21,025 rules, and double rotations, under 24 ways of using the table, no secret word shorter than 22 letters survives, and no puzzle word fits the forced letters.

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.

  1. Pick a rule: start at 5, step 7. The first plaintext letter comes from position 5, the second from 12, the third from 19, and so on, wrapping past 97.
  2. Position 22 of the plaintext, the E of EAST, is the 22nd letter read, so it comes from 5 + 7 x 21 = 152, which wraps to 55.
  3. The ciphertext at 55 is therefore the enciphered E; under the sculpture's alphabet and Vigenere the key letter that sends E there is forced.
  4. Do this for all 24 known letters to get 24 forced key letters at 24 positions.
  5. Two positions whose distance is a multiple of p share a key slot at period p; if their forced letters differ, p is dead.
  6. The smallest live p is the shortest secret word the rule allows.
  7. Slide the puzzle's words across the survivors' pattern.
the worksheet: the first steps by hand, on squared paper
THE LEDGER FOR START 5, STEP 7: WHERE EACH KNOWN LETTER COMES FROM, AND THE KEY LETTER IT FORCESplaceknown5 + 7(p - 1)wraps tocarvedkey letter22E15255UD23A15962FO24S16669TX25T17376GC26N18083CJ64B44658AZ65E45365YLCount on by the step from thestart, wrapping past 97, to findthe carved letter that holds eachknown letter.Each pair of known and carvedletters forces one key letter onthe sculpture's table. Two placesa key length apart must force thesame letter.The experiment does this for everystart and step: 21,025 rules.
4

Worked example, generated

The same steps run by the code on the real letters.

route: K3 rule, start 5, step 7 (plaintext[i] comes from position 5 + 7 i, 0-indexed); convention kry/kry/kry Vigenere

plain pos  letter  lands at  cipher there  forced key letter
       22    E        31         G              Y
       23    A        45         E              T
       24    S        59         W              L
       25    T        73         P              Z
       64    B        37         W              I
       65    E        51         T              N
       66    R        65         Y              R
       67    L        79         Z              B

consecutive crib letters land 7 apart, so period 7 and period 1 need their forced letters equal.
5

Hypothesis, for a cryptanalyst

Family, key space, the constraint under test, and the clues it reads.

Sanborn, 2003, secondhand (clue C2): surprised no one had tried 'recovering the original matrix and running it through all the possible shifts'. K3's matrix is known: Stehle's rule reads plaintext[i] = ciphertext[(191 + 192 i) mod 337] (data/k3.json), and the NSA described it as a double rotation of a grid. Reading (a): K4 is the same matrix with a different shift. For n = 97, prime, every pair (a, b) with b in 1..96 is a valid rule, 9,312 routes, each applied forward and inverse; add the double rotations, write in rows of w1 and read columns bottom-up, then again with w2, for w1, w2 in 2..50. Each route goes before a periodic tableau with separate alphabets, as in 0006, and the cribs force key letters at their new positions; the divisor rule gives the smallest live period. Live periods up to 30 are matched against the vocabulary and K1-K3 windows. Bounds: complete for this route set, 24 conventions, periods 1-97.

6

Prediction

Written before the run; the result decides it.

As in 0006, the smallest live period clusters between 24 and 50. No route allows a period of 15 or less. No forced fragment matches a vocabulary word or K1-K3 window; zero candidates pass.

7

Method

  1. Build the routes: plaintext[i] = inter[(a + b i) mod 97] for a in 0..96, b in 1..96, forward and inverse; double rotations of a grid for widths 2-50 twice.
  2. For each route, send each crib position to its intermediate position.
  3. For each convention, force the key letter from the crib plaintext and the ciphertext at the destination.
  4. Period p is dead if it divides a distance between two destinations with different forced letters; the smallest live p is the route's score.
  5. For live p up to 30, match the forced fragment against the vocabulary and K1-K3 windows; decrypt matches and submit.
8

Result and bounds

The harness's record: candidates, gates, loss, and the negative result this experiment leaves behind.

0candidates returned
0passed every gate
0/24known letters, best
–lowest loss
nonechampion before

504600 route/convention pairs; 0 forced-fragment matches submitted. K3 rule verified against data/k3.json inside run.

What was searched, as the harness recorded it.

family
K3 affine rule with all shifts, or double rotation, then periodic tableau (separate alphabets)
routes
21,025
route convention pairs checked
504,600
route set
plaintext[i] = inter[(a + b i) mod 97], a in 0..96, b in 1..96, fwd and inv; double rotation w1, w2 in 2..50
conventions
24
periods
1-97 by divisor rule
key pool
vocabulary words and K1-K3 windows of length P for live P <= 30
min live period histogram
22
4
23
22
24
492
25
1,276
26
6,644
27
8,744
28
8,430
29
9,646
30
11,430
31
24,652
32
25,370
33
38,770
34
33,660
35
29,338
36
29,950
37
15,774
38
29,406
39
27,084
40
17,890
41
18,424
42
16,512
43
24,298
44
18,380
45
10,120
46
16,502
47
25,208
48
7,740
49
5,308
50
8,486
51
2,816
52
4,466
53
1,052
54
820
55
1,990
56
1,158
57
878
58
1,112
59
1,268
60
1,194
61
922
62
1,162
63
834
64
1,010
65
938
66
1,090
67
814
68
890
69
716
70
732
71
658
72
634
73
796
74
482
75
528
76
626
77
578
78
710
79
384
80
438
81
404
82
528
83
350
84
186
85
248
86
212
87
212
88
210
89
488
90
62
91
80
92
14
93
16
94
334
fifteen shortest
  1. min period
    22
    route
    K3 rule a=24 b=2 fwd
    convention
    std/std/std/beau
  2. min period
    22
    route
    K3 rule a=24 b=2 fwd
    convention
    std/std/kry/beau
  3. min period
    22
    route
    K3 rule a=85 b=49 inv
    convention
    std/std/std/beau
  4. min period
    22
    route
    K3 rule a=85 b=49 inv
    convention
    std/std/kry/beau
  5. min period
    23
    route
    K3 rule a=60 b=8 fwd
    convention
    std/kry/std/beau
  6. min period
    23
    route
    K3 rule a=60 b=8 fwd
    convention
    std/kry/kry/beau
  7. min period
    23
    route
    K3 rule a=53 b=12 inv
    convention
    std/std/std/beau
  8. min period
    23
    route
    K3 rule a=53 b=12 inv
    convention
    std/std/kry/beau
  9. min period
    23
    route
    K3 rule a=41 b=85 inv
    convention
    std/kry/std/beau
  10. min period
    23
    route
    K3 rule a=41 b=85 inv
    convention
    std/kry/kry/beau
  11. min period
    23
    route
    K3 rule a=36 b=89 fwd
    convention
    std/std/std/beau
  12. min period
    23
    route
    K3 rule a=36 b=89 fwd
    convention
    std/std/kry/beau
and 3 more
vocab matches
0

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.
view on GitHub