Experiment 222
A6.1 (Stage 2, step 4, issue #84), the second group of sampled verdicts. 0034 (one periodic key on the standard and KRYPTOS tables, 24 conventions, periods 27-29 and 53-95 every third), 0035 (27-29, one side keyword-mixed from the vocabulary) and 0048 (27-29, both sides keyword-mixed) completed the free key slots by a quadgram climb and compared K4's best with shuffled K4 by quadgram. Their copies were never kept, and their runs stopped on the clock, so they cannot be replayed letter for letter; but 0216 ran each frozen search unchanged on 10 fresh plain shuffles (tools/a13_check.py pass1-regen, seed 1300) and kept every reading (reports/a13-structure-data/pass1_copies.json). Here K4's saved readings and those copies' readings go to the judge in one blind call per experiment. 0037's sampled triples are not rejudged: F6.3b's close rests on 0158's exact count.
What the method does with the 24 known letters.
Three old experiments filled in long keys so that the carved lettersthe carved letters The 97 letters cut into the copper at the CIA, shown at the top of every experiment page. read as much like English as possible, and compared the result with the same trick on scrambled copiesa shuffled control The same experiment run on the carved letters shuffled into a random order. using a simple letter-statistics score. That score is poor at telling damaged English from English-looking junk. Here the judgethe judge A language model that scores how much of a reading is English, from 0 to 100, allowing misspellings and X between sentences., a language model that reads one text at a time without knowing where it came from, reads the carved letters' readings and the copies' readings.
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.
Three old searches filled in long keys to make K4 read like English and compared it with scrambled copies by letter statistics. The judge reads for words instead.
K4's 24 saved readings and the same searches' readings of 10 copies each, 264 texts in all, judged blind.
K4 above every copy only at 0034's lengths 27-29, by 8 against 7 out of 100 (chance 0.36 such cells): a candidate, replicated in 0224/0225.
What this experiment was built on, from the lineage page: 14 experiments before it, back to the root of its line. Its number is in bold.
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 candidate with most known letters right; none passed every gate.
in: the copper, always the same
OBKR ends the line that finishes K3, then three lines of 31
out: the candidate
checked: the known letters in place, and the reading turns back into the copper
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 -3.28; English text scores -3.6 | 0.00 |
| Key | 461 bits to write it down (key words are cheap, arbitrary letters 4.7 bits each); each 60 beyond 60 costs a point | 6.68 |
| Loss | of 240, the worst possible | 6.68 2.8% |
Redo the method on a few letters with pencil and paper.
The same steps run by the code on the real letters.
0035's reading of K4 (period 27, plain std, cipher HILL/keyword, vig; 3 free slots climbed):
reading QGFMJOYSTOAEUHISNORDFEASTNORTHEASTSDKZSIZKIENYYISBUIIAWIVYLBOMOBERLINCLOCKFFEMXURMHAWYLVEERAJBUXH
judge sees QGFMJOYSTOAEUHISNORDF#############SDKZSIZKIENYYISBUIIAWIVYLBOMO###########FFEMXURMHAWYLVEERAJBUXH
the judge reads it among 30 readings of 10 scrambled copies by the same search, one at a time, unlabelled.
Family, key space, the constraint under test, and the clues it reads.
A6 (knowledge/12-assumptions.md); C27(c) and C5 for the tables (the keyword-mixed alphabets of 0035 and 0048), C19 for the judge's tolerance of X stops. Cells: 0034 at 27-29 (3 readings per text), 0034 at 53-95 (15 readings; already blocked by 0136's bound, judged for the record), 0035 at 27-29 (3), 0048 at 27-29 (3). Statistic per cell: the judge's (j3, Sonnet 5.5) best score over a text's readings; K4 counts only strictly above all 10 copies' bests; None as not found. Every reading of every text is sent (no quadgram cut, the same for K4 and copies). Four cells, chance about 4/11 = 0.36 cells with K4 above every copy; a flag needs p < 0.05/4, 80 copies, so a K4 above every copy here is a candidate to replicate on that cell against fresh copies, not a result.
Written before the run; the result decides it.
K4 above every copy in no cell (chance 0.36). The copies' best judge scores are 20-50 at 53-95, where the climb fills 47-71 free letters with English-like fragments, and under 20 at 27-29, where only 3-5 letters are free. Zero candidates pass gate 4.
The harness's record: candidates, gates, loss, and the negative result this experiment leaves behind.
All 24 candidates, lowest loss first.
saved reading of 0048: period 29, plain LUCID/keyword, cipher WONDERFUL/keyword, beau; 5 free slots climbed
saved reading of 0048: period 27, plain WEBSTER/keyword, cipher WORLDCLOCK/keyword, beau; 3 free slots climbed
saved reading of 0035: period 28, plain WILLIAMWEBSTER/keyword, cipher std, beau; 4 free slots climbed
saved reading of 0048: period 28, plain SHADOW/keyword, cipher LODESTONE/keyword, varbeau; 4 free slots climbed
saved reading of 0035: period 29, plain LANGLEY/keyword, cipher std, beau; 5 free slots climbed
saved reading of 0035: period 27, plain std, cipher HILL/keyword, vig; 3 free slots climbed
saved reading of 0034: period 29 key HYHMKBNENTXBQRMAEROOOXYHANZEF (5 free slots climbed); kry/kry/std/beau
saved reading of 0034: period 28 key NZVDMORFNOEEADIAGKSSSUELRXLO (4 free slots climbed); kry/std/std/vig
saved reading of 0034: period 56 key BFFCTASVNMWWAXSAUQSJPGWPJDPMNBFXOMFSIWDMFVJTLJFMOEMVEAHI (32 free slots climbed); kry/
saved reading of 0034: period 53 key PRPUBCROXCHYSSRSHDNLQCECHPTSDQDDDEHDKTQUHQMEEHLMPTBKH (29 free slots climbed); std/std
saved reading of 0034: period 68 key SDHLGKTHUVFNRRDEIXWHPREZYPYXXSYDKZSSVSGKUTHKTWVLCAFONAADRVOJCODFQXDE (44 free slots cl
saved reading of 0034: period 27 key EFHYHMKUKBNENTXBQRMAEXYHANZ (3 free slots climbed); kry/kry/std/beau
saved reading of 0034: period 65 key TGNWRJLFKIYNHDJRTJHPXWXAKGZTOAXAFDJDFWXQRUGDKOMSJLBTEWZZQKPTPWCRG (41 free slots climb
saved reading of 0034: period 74 key LGZVHKFBYMIMUHPSIHZMLQTELWEHGZUPFHUIBBFTXDHEKZNUABBJSOGDXEPUSRFOGHTTKPBKSD (50 free sl
saved reading of 0034: period 71 key KSDEGWSDMKXVMAYZNVJUIQTELWEHGZUPFHUFMTEZYWBPEPUHJVTJPMDVTVBWFQUOGHTTKPB (47 free slots
saved reading of 0034: period 77 key HSUPMCUYLTXJVEGYTUVDPMCLWNQHMHLKFGFKQKXYGDAGNLNRQWVRTYNCMWQEKMCBYHITCDFYZDURP (53 free
saved reading of 0034: period 59 key IYIFUCJFBFUJDXDKGVFJYJMWZCUJURLLXVUFSBCVWRNZBZMFKHTIHKGANRY (35 free slots climbed); k
saved reading of 0034: period 62 key FOGHTTKPBKSDJZDKAFFBLQTELWEHGZUPFHDOEJCLOOLEEKMSFEKOVKEXIVEASR (38 free slots climbed)
saved reading of 0034: period 80 key XXZVLYUUDKQKHVOGLYXCTBKVSBPCXFCTACADMYQSJQCSAZOSTDQFWDKDOWQWJXPLRCFPLJIHUATZZZOS (56 f
saved reading of 0034: period 83 key JSRDFXJVYXNXFNCNCSAFWSJRXTUOSORAMNCAZEZBOGLYRONFQMDORNNDEXUPZOFICOPEJKMCZKGEUSUCPSE (5
saved reading of 0034: period 86 key IZHXAZWDYMNTQMMIWYHMXZPBYXYCCUYQABQYIZDWIOIAPUWFVEZRDSOKLHILUZROGCQPUQMJNAXUQJIHJFLQZY
saved reading of 0034: period 89 key NACPMJYYGVZKPKXKWQZMXQRTPHIUUARAYZSHZSHZXYXKDEQHKALUSKZBPUATKMVUGUHMRUKAFKAEONSFIHNKGN
saved reading of 0034: period 92 key KVUKDZQJQDDANGRPTOAWTRDUMRIYWOYNKYRJFXJMOHYMKZHSNVPDNXEUVQJVCWIELYOIECBAQKDASNSHKYXCTQ
saved reading of 0034: period 95 key FALPWPOZKOFGDGYGYMDMJMCLWNQHMHLKFGXKQHZRHUEXABHQDWGHKKCPTKAYEAZBYHITCDFYZDATQMCOKGZPLS
0034 P53-95: K4 30, copies best 45 (rank 8 of 11); 0034 P27-29: K4 8, copies best 7 (rank 1 of 11); 0035 P27-29: K4 6, copies best 12 (rank 8 of 11); 0048 P27-29: K4 12, copies best 12 (rank 6 of 11)
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.