“The last corner of family 1: both alphabets rebuilt from the puzzle's words, key completed at periods 27 to 29, sampled”
Experiment 35 mixed one alphabet at a time. Here both the message alphabet and the carved alphabet are keyword mixings of vocabulary words, 61 x 61 pairs, drawn in random order with a fixed seed until the budget runs out; the record says how many pairs were reached. Periods 27 to 29, three restarts.
Experiments 34 and 35 filled in the few unknown letters of a secret word of length 27 to 29 and found only nonsense, with 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. and with tables rebuilt from one puzzle word on one side. The last version has both sides of the table rebuilt from puzzle words. There are too many pairs of words to finish in one sitting, so we draw pairs at random for half an hour and report how far we got.
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 key
A repeating secret word of 29 letters on two tables built from puzzle words (here LUCID for the message and WONDERFUL for the copper). The known letters fix 24 slots; the climb tries every letter in the other 5.
the scale
The best text at each period sits with what the same climb finds on a shuffled copper, far from English at about -3.6.
the bound
All 3,721 table pairs at periods 27 to 29, 33,489 climbs: the readings that pass every gate do so by construction and read as noise. Periodic keys at these lengths are closed for vocabulary tables.
2
Best candidate against the known letters
Laid out as the copper carries K4: OBKR ends the line that finishes K3, then three lines of 31. Left, as carved, the 24 known places in accent. Right, the candidate: a known place it gets right is an accent cell, one it misses is outlined. This is the lowest-loss reading that passed the original gates.
in: the copper, always the same
as carved, 97 letters; the 24 known places in accent
ECDMRIPFEIMEHNLSSTTRTVDOHW?OBKR
UOXOGHULBSOLIFBBWFLRVQQPRNGKSSO
TWTQSJQSSEKZZWATJKLUDIAWINFBNYP
VTTMZFPKWGDKZXTJCDIGKUHUAUEKCAR
OBKR ends the line that finishes K3, then three lines of 31
out: the candidate
every place read; known places in accent, misses outlined
ECDMRIPFEIMEHNLSSTTRTVDOHW?JGAF
EEFGHJCJOUSTOLITIEASTNORTHEASTE
HHUAODBULYMENDTRYFFLAWPKASEDBER
LINCLOCKITHNDPRGADTYGABNDEDOZMS
checked: the known letters in place, and the reading turns back into the copper
3
How the score is made
The loss of the best candidate, part by part. Zero is solved.
Known letters missed
0 of 24, ten points each
0.00
English
the 73 free letters score -6.14; English text scores -3.6
2.54
Key
150 bits to write it down (key words are cheap, arbitrary letters 4.7 bits each); each 60 beyond 60 costs a point
1.50
Loss
of 240, the worst possible
4.04 1.7%
All 3 candidates that passed every gate, lowest loss first.
rank
missed
English
key
loss
share
1
0
2.54
1.50
4.04
1.7%
best
2
0
2.75
1.35
4.10
1.7%
3
0
2.92
1.43
4.35
1.8%
4
Walkthrough, by hand
Redo the method on a few letters with pencil and paper.
Build the alphabet from PALIMPSEST: PALIMSETBCDFGHJKNOQRUVWXYZ, and use it for the message; keep KRYPTOS for the carved letters; Vigenere.
Period 27: position 22 is E, index 6 in the PALIMPSEST alphabet; carved F is index 11 in KRYPTOS; the key index is 5.
Do this for all 24 known letters; positions 22 and 49 share slot 21, but 49 is unknown, so no clash.
Three slots stay free.
Try every letter in each free slot in turn, decrypting the whole text each time, and keep the one that makes the 73 unknown letters read most like English.
Compare the best text's score with English, about -3.6, and with what the same game gives on shuffled letters, about -6.8.
the worksheet: the first steps by hand, on squared paper
5
Worked example, generated
The same steps run by the code on the real letters.
plain alphabet PALIMPSEST/keyword PALIMSETBCDFGHJKNOQRUVWXYZ; cipher alphabet KRYPTOS; Vigenere; period 27; 3 free slots climbed
key indices [10, 20, 6, 7, 25, 1, 25, 11, 14, 11, 22, 10, 20, 1, 14, 9, 23, 21, 20, 11, 18, 6, 16, 22, 15, 4, 3]
text VJUUWMZUZIZTJSOBQOJXSEASTNORTHEASTUNGBDGKEXCNEJTOOBAWJFNCPYJUAUBERLINCLOCKSTUMDUAEKMBJZWITWZAEBRQ
English of the free letters -8.243 (English about -3.6, noise about -6.8)
6
Hypothesis, for a cryptanalyst
Family, key space, the constraint under test, and the clues it reads.
Experiments 34 and 35 found the key completion at periods 27-29 reads as noise under every pair with at most one keyword-mixed side. The remaining corner is both sides mixed: 3,721 pairs x 3 modes x 3 periods, about 33,000 climbs at 0.1 s each with three restarts, more than the budget allows, so pairs are visited in a fixed random order and the run stops at the clock; coverage is reported. A true periodic cipher on such a pair would give a text near -3.6 at its period; shuffled K4 gives about -6.8. Bounds: a stated fraction of the pair space, not exhaustive; the seed makes the sample reproducible and a later run can continue it.
7
Prediction
Written before the run; the result decides it.
Within the budget about a third of the pairs are reached. No pair and period gives a text better than -5.5 per quadgram; the best is within 0.4 of shuffled K4. Candidates pass the gates by construction but none reads as English or beats the current champion's fitness.
8
Method
Build the alphabet pool: standard, KRYPTOS and the keyword mixing of each vocabulary word.
For each pair with one side standard or KRYPTOS, each mode and each period 27-29, force the 24 key slots from the cribs; skip the pair if two cribs disagree on a slot.
Climb the free slots by coordinate descent on the English score of the 73 non-crib letters, five restarts.
Keep the best text per period and submit it.
Repeat on one shuffle of K4 for calibration.
9
Result and bounds
The harness's record: candidates, gates, loss, and the negative result this experiment leaves behind.