• ▸We believe the solution is present within these families4 open, 32 blocked, 8 noted, 139 closed

  • ▸Excluded by the known letters alone, before any search4 closed

    • Pure transposition: the known letters need three Es, the copper has twoclosed · exact · 51
    • One substitution alphabet: the same carved letter stands for different known lettersclosed · exact
    • Rotor machines: K encrypts to K, S to Sclosed · exact
    • Systems with 25 or fewer symbols (Playfair, bifid, ADFGVX): K4 uses all 26 lettersclosed · exact · 52
  • ▸1stFamily 1: One keyed stage, no move4 open, 12 blocked, 4 noted, 38 closed

    • ▸A repeating key on the table3 open, 3 blocked, 1 noted, 8 closed

      • Standard or KRYPTOS alphabets, key lengths 1-26 and 30-52closed · exact · 5
      • ▸Key lengths 27-29 and 53+ on alphabets from the puzzle's words; the completed key reads as noise1 blocked, 1 closed

        • Lengths 27-29 and 53: a planted text is told from its copies (C-F1-2, C-F1-2b)closed · sampled · 34 35 48 53
        • Lengths above 53: a planted text at 80 is not told from its copies (C-F1-2d); 0136's boundblocked · sampled · 34
      • Alphabets from any dictionary word, key lengths 1-26closed · exact · 109
      • Alphabets one swap or move from KRYPTOSclosed · exact · 71
      • ▸F1.1 One alphabet in an unknown order on both sides (Quagmire III), key lengths 3-26; Beaufort 23 and 26 rerun against 100 own copies with MIX8m and the judge (F1.12): K4 inside on all three2 open, 1 blocked, 2 closed

        • Added form (Quagmire III proper, and variant Beaufort), lengths 1-26: S under S and K under K force key 0 in slots another known letter sharesclosed · exact · 119
        • Beaufort form, lengths 1-22, 24, 25: no alphabet fitsclosed · exact · 119
        • Beaufort form, length 23 (7 checks): design 2 (F1.5) puts K4 9th of 13 among copies of itself; planted texts beat all their own copies 6 of 8 (0139, again 0228), Sanborn-style plants 10 of 12 (0226), 6 of 12 (0228), 16 of 24 (0229): pooled 32 of 48, under the bar; clean pooled 33 of 40: reopenedopen · sampled · 119 120 139 140 226 228 229
        • Beaufort form, length 26 (1 check): K4 12th of 13 among copies of itself; power 5 of 6 (0140), redrawn 13 of 24 clean and 17 of 24 Sanborn-style (0229), under the bar: reopenedopen · sampled · 119 140 142 229
        • Masked messages at the sampled lengths (A6.3): reversed found 7 of 8; an X every kth letter 5 of 8, the message in every nth place 3 of 8, under the 6 of 8 bar: not decidedblocked · sampled · 227
      • ▸F1.2 Two alphabets in unknown orders (Quagmire IV), key lengths 3-26; length 13 power remeasured on 48 plants (D1b: 16/24 clean, 16/24 Sanborn-style at ten runs) and rerun against 100 own copies (F1.12): K4 last by quadgrams, inside on all readers1 open, 1 blocked, 1 noted, 2 closed

        • Lengths 1-7, 9-12, 14, 15, 17, 18, 21, 22, 25: no pair of alphabets fitsclosed · exact · 123
        • Lengths 8, 16: design 2 on two alphabets puts K4 2nd and 7th of 13 among kept-place copies; power 4/4, 4/5 on clean Carterclosed · sampled · 143 145 146
        • Length 13: K4 13th of 13 (0144), power 5/5 on clean Carter (0143) but 9/12 on Sanborn-style plants, under the 4/5 bar (A6.2, 0226); a second draw 2/5 clean, 8/12 messy (0228): reopenedopen · sampled · 143 144 226 228
        • Lengths 20, 23, 24, 26: above 0136's bound, only a better English model could decideblocked · exact · 123 136
        • Length 19: a pair fits, 26 of 1,000 coppers do (R under P at 28 and 66, 38 apart)noted · exact · 123 124
    • A progressive keyclosed · exact · 11 42 155
    • ▸A self-feeding key (autokey)2 blocked, 3 closed

      • Ciphertext autokey, any lag, on 1,092 keyword tablesclosed · exact · 2 25 64 118
      • Plaintext autokey, lags 1-52closed · exact · 3 12
      • ▸Plaintext autokey, lags 53-96: the known letters force too little2 blocked

        • Lags 85-96: open bits above what 73 letters of English carry at Shannon's lowest entropy; no English test can decideblocked · exact · 136
        • Lags 53-84: above the quadgram scorer's 130 bits; only a better English model could decideblocked · exact · 136
      • Autokey plus a repeating keyclosed · exact · 66 67 73
    • ▸A running key from a text3 closed

      • K1-K3, the sculpture's texts, Carter, Sanborn's lettersclosed · exact · 4 17 21 45 47 78 98 117
      • An unknown English text, on standard, KRYPTOS or puzzle-word alphabetsclosed · sampled · 9 10
      • F1.3 An unknown English text, on any dictionary-word alphabet; Sanborn-style running keys found 12 of 12 by quadgrams and the judge (A6.2, 0226)closed · sampled · 125 226
    • ▸A key schedule worked by hand1 open, 2 blocked, 3 closed

      • The Berlin clock, the Carter datesclosed · exact · 79 38
      • ▸F1.4 Other hand schedules (C20(c)-(e))next1 open, 2 blocked, 2 closed

        • The tableau rows as repeating keys ("layer two", C20(d)): 6 of 24 known letters, coppers 5-8closed · exact · 126 128
        • A key word stepped by skip, stutter, back and forth, triangle or square numbers, or restarted at each carved line, on standard or KRYPTOS tables: no setting fitsclosed · exact · 127 128
        • The same rules on free alphabets: consistent at the coppers' rate; a decider exists (F1.5) but about 1,900 settings at 20-30 min each: blocked by costblocked · exact · 127 128
        • K5 as the key (C20(e)): no textblocked
    • ▸F1.5 A decider for cells the known letters cannot decide (the shared blocker of F1.1, F1.2, F1.4's free alphabets, autokey lags 53-96, F6.1, F6.2, F8): bound each cell by key bits against English redundancy, then a climb that passes 0120's power bar on planted texts (P6)2 noted, 3 closed

      • Step 2, design 1 (check-keeping climb, compiled): plants beat K4's copies 10 of 12 but only 6 of 12 their own: power came from letter counts, failsclosed · sampled · 137 138
      • Step 2, design 2 (parallel tempering in the check-keeping set): plants beat their own copies 6 of 8 at 23, 5 of 6 at 26, 4/4-5/5 with two alphabets; decides F1.1 and F1.2's cellsclosed · sampled · 139 140 143 145
      • Design 2 against its budget (D1): one full run matches ten at 23, nothing under nine at 13; the ten-run power redrawn lower than recorded (messy 6/12 at 23, clean 2/5 at 13)noted · sampled · 228
      • One free alphabet with 24 plants per class (D1): power does not rise with runs (L23 clean 21-23, Sanborn-style 14-16; L26 11-18, 11-17 over r = 1-10); strict rule not metnoted · sampled · 229
      • Step 1, the bound: the quadgram scorer extracts 1.78 bits per letter (130 bits in 73 letters); one free alphabet at every length, two to 16 (14 with a move), Hill with any key decidable in principle; two alphabets from 19 (15 with a move) and autokey 53-84 only with a better model; autokey 85-96 by noneclosed · exact · 136
    • ▸F1.6 One carved letter wrong, missing or extra, through the exact unknown-alphabet tests4 closed

      • One and two free alphabets, periodic, 604 variants: at the coppers' rateclosed · exact · 147
      • K1/K2's system revived at length 19 by a slip at 33: decider power 3 of 5, K4 7th of 13: sampled at power 3/5 (D3)closed · sampled · 148
      • Gromark (all free-alphabet rules and bases), 605 variants: no fitclosed · exact · 150
      • Trifid and Hill, every single-letter error anywhere (5,044): no cube; Hill at the coppers' rateclosed · exact · 151 152
    • ▸F1.7 Autokey on one unknown alphabet, with no move, then after each move1 blocked, 2 closed

      • No move: dead lags closed; fitting lags at the coppers' rateclosed · exact · 149
      • Fitting cells (130 under full chain equations; 3 of 0149's dead) judged by English against fitting copies and 10 stand-ins: K4 p 0.64; plants 12/48 (F1.13, a rerun with a better climb, blocked: no decider found, 2026-11-04)closed · weak · 256
      • The move-first version: blocked by costblocked
    • ▸F1.8 Carter's book as a running key on an unknown alphabet1 blocked, 1 closed

      • One alphabet, every offset both ways: no fit on K4 or copiesclosed · exact · 154
      • Two alphabets, any running key: almost no checks; undecidable by the known lettersblocked · exact · 154
    • ▸F1.9 The per-letter progressive key; its periods 26-30 against coppers1 noted

      • Period 34 (K4 40 and 20 consistent pairs against copy maxima 10 and 7), inside the fixed statisticnoted · exact · 155
    • F1.10 Friedman's keyword-only key-advance schedulesclosed · exact · 156
    • F1.11 A split key phrase: one part mixes the plain alphabet, the other the cipher alphabetclosed · exact · 157
    • ▸A1.2 One, two or three of the 24 known letters slipped (A1, Stage 2 step 2), periodic key1 blocked, 3 closed

      • The sculpture's tables, lengths 1-26: nothing fits K4 below 25 even with 3 slips; 24-26 fit everywhere; the judge puts K4's best reading at 3 against the copies' 8closed · exact · 207
      • One free alphabet: dead at 1-7 and 9 with up to 3 slips; plants at 7 foundclosed · exact · 208
      • Two free alphabets: dead at 1-3 and 9 with up to 3 slips; plants at 3 and 9 foundclosed · exact · 209
      • Free-alphabet lengths where K4 and the copies fit with 2-3 slips (one alphabet 8, 10-26; two 4-8, 10-26): too short to decide k slips, the reading undeterminedblocked · exact · 208 209
    • ▸A1.3 Running keys from the named texts with slips at the known places1 blocked, 2 closed

      • K1-K3, the sculpture's texts, Carter, Sanborn's letters on standard, KRYPTOS and keyword tables, up to 7 slips: no setting on K4 or any copyclosed · exact · 210
      • The sculpture's texts and Carter plus a constant shift, with or without every type-1 move, up to 3 slips: noneclosed · exact · 211
      • On a free alphabet (0098, 0154): one slip lets hundreds of settings fit on every copy, reading undetermined: too short to decide k slipsblocked
    • ▸A1.4 Autokeys and progressive keys with slips or one lost step2 closed

      • Ciphertext autokey on 487,734 table setups, 2 slips or a lost step: nothing below lag 68; progressive key, 2 slips: nothing below period 25closed · exact · 212
      • Plaintext autokey, lags 1-52, 2 broken links or a lost step: nothing below lag 24; lags 26-29 read blind as copiesclosed · exact · 213
    • ▸A1.5 A move, then the table, with two slips1 blocked, 1 closed

      • 0110-0114's routes and the type-1 library (1.85M landings), 8 tables, keys 1-26: shortest fit 18 after, 16 before, as copies; readings judged as copies'closed · exact · 214
      • A free alphabet after a move with slips (0129): fits at the copies' rate from length 6 without slips; too short to decide k slipsblocked
  • ▸2ndFamily 6: A move, then the table4 blocked, 2 noted, 34 closed

    • ▸One move, then a repeating key on standard, KRYPTOS or word alphabets6 closed

      • Columnar (plain, keyed, double, Myszkowski, AMSCO), rail fenceclosed · exact · 6 26 58 84 85 110 111 112 113 114
      • K3's rule, quarter turns, grid paths, the ? as a 98th symbolclosed · exact · 8 41 82 83 86 99
      • Key on both sides of the move, the same table both sidesclosed · exact · 80
      • F6.22 AMSCO at key length 23: above the coppers in 0114; against 200 fresh shuffles K4 21, 4 of 200 at or above (p 0.025, 0.075 over three lengths), no forced key a named passageclosed · exact · 114 0263; chased in F6.22b
      • F6.22b The forced period-23 keys read for English against the shuffles'closed · sampled, plants 20/20 · 0264 (flawed score) 0265: K4's best key -11.9 bits below random; p 0.03 on more settings; chased in F6.22c
      • F6.22c The same against a count-matched nullclosed · sampled · 0266: p 0.06 per setting p 0.84
    • One move with an autokey, running key or clock scheduleclosed · exact · 63 81 90 91
    • One carved letter wrong, missing or extra, through every moveclosed · exact · 93 94
    • ▸Three stages3 closed

      • Three clockwise quarter turns, every widthclosed · exact · 83
      • ▸F6.3 Mixed three-stage routes2 closed

        • Four column readings (turn either way, rows-then-columns-down, its reverse), every pair at widths 2-96 and every triple at 2-24: nothing live below key length 20closed · exact · 134
        • F6.3b Mixed triples at widths 25-96: 105.5 million routes, nothing live below 18, at the coppers' rateclosed · exact · 37 158
    • ▸Alphabets in unknown orders after or before a move2 blocked, 1 noted, 2 closed

      • ▸F6.1 One unknown alphabet (Quagmire III)1 blocked, 1 closed

        • Every type-1 move, either order, added or Beaufort form, key lengths 1-5: no alphabet fitsclosed · exact · 129 130
        • Lengths 6-26: settings fit at the scrambled-copper rate; only an English climb could judge them (0120)blocked · exact · 129 130
      • ▸F6.2 Two unknown alphabets (Quagmire IV)1 blocked, 1 noted, 1 closed

        • Every type-1 move, either order, key lengths 1-2 (and 3 with the table first): no pair fitsclosed · exact · 133
        • The rest: fits at the scrambled-copper rate (exact to 10, per-slot above); only an English climb could judgeblocked · exact · 133
        • Move first, key length 3: K4 at the top of its coppers in 0097, 0100, 0133; p 0.04-0.10 (0102)noted · exact · 100 102 133
    • ▸F6.4 Padding at each grid stage; nulls dropped between grids1 closed

      • Blanks in any of eight places per grid, dropped or kept as dummies, one stage (widths 2-96) or two (2-48), periodic key 1-26 on standard/KRYPTOS tables: nothing live below 21closed · exact · 135
    • F6.5 Direction words as routes through the grid ("east, then northeast"), then the tableclosed · exact · 159
    • F6.6 Move, change, move again (844-move library, all pairs)closed · exact · 160
    • F6.7 Change, move, change on two different tablesclosed · exact · 161
    • F6.8 Three-stage routes with a keyword columnar stageclosed · exact · 162
    • F6.9 Both autokeys with 0084's width-7 double columnar routesclosed · exact · 163
    • F6.10 Blank cells chosen by a key, then the tableclosed · exact · 165
    • F6.11 Keyword-disrupted and triangle columnar transpositions, then the tableclosed · exact · 164
    • F6.12 EVA's double triangle transposition, then the tableclosed · exact · 166
    • F6.13 Chiffre XIII's keyword-shaped five-row mask, then its period-5 table (parametric masks; the documented rule not in hand)closed · exact · 166
    • F6.14 Nihilist transposition on a 10 x 10 square with three blanksclosed · exact · 167
    • F6.15 YAR's "1-2-3" spacing as an ordering, then the tableclosed · exact · 166
    • F6.16 Interrupted-key columnar transposition, then the tableclosed · exact · 164
    • F6.17 Trapezoid transposition, then the tableclosed · exact · 164
    • F6.18 Letters reshuffled inside each row at widths other than 7, then the tableclosed · exact · 165
    • F6.19 The French diagonal method, then the table (key-named diagonals)closed · exact · 165
    • ▸F6.20 A 10x10 turning grille (Fleissner), then or before the table2 blocked, 2 closed

      • Copied out by rows, every card, free keys 1-6 and 200,563 key words: no card fitsclosed · exact · 199
      • Copied out by columns, every card, free keys 1-4 and the key words: no card fitsclosed · exact · 200 201
      • Free keys of 7-26 letters (rows) and 5-26 (columns): 10^8-10^9 nodes per settingblocked · exact · 199 201
      • Fixed stencils (the Abwehr's 13x19 card): no rule for the holesblocked
    • ▸F6.21 A deliberate slip in a keyword columnar (one column out of place, or two swapped), then the table1 noted, 3 closed

      • Around the dictionary keyword columnar, the same word twice, the interrupted key: at the copies' rateclosed · exact · 202
      • Around the trapezoid and the poem code: at the copies' rateclosed · exact · 203
      • Around the row reshuffle: 6 cells above every copy (0203), 1 against 40 fresh copies (0204): chanceclosed · exact · 203 204
      • A21 of the row reshuffle, carried by rows of 9noted · exact · 204
  • ▸3rdFamily 12: Stein's way: look, quantify, then crib4 blocked, 3 closed

    • F12.1 The 97 letters and the "?" checked against images of the copperblocked
    • F12.2 K4's statistics by segment and width against shuffles; a planted period-7 key is not seen (C-F12.2)blocked · sampled · 168
    • F12.3 Stein's I.C. peak at 13 in K1 and K2: real or chanceclosed · sampled · 169
    • F12.4 A pencil session: every pattern noted before any code; planted patterns not seen (C-F12.4)blocked · sampled · 170
    • F12.5 Parts with their own keys, split where the statistics breakclosed · exact · 171
    • F12.6 Probable words through the cells the known letters cannot decide; a plant scores below its copies (C-F12.6)blocked · exact · 172
    • F12.7 Stein's K1-K3 method as such: column shifts scored by log frequencies, then a cribclosed · by subsumption · 5
  • ▸4thFamily 3: A two-part key with a combiner11 closed

    • Two words added: both from the puzzle, or one from the dictionaryclosed · exact · 14 43 59 115
    • A word plus digits from the storyclosed · exact · 24 38
    • A word plus the misspelt or raised lettersclosed · exact · 61
    • A column word plus a row keyclosed · exact · 87 88
    • F3.1 Combiners other than additionclosed · exact · 173 174
    • F3.2 Number keys from the Morse and the plate layoutclosed · exact · 175
    • F3.3 Direction words as numbers: the bearing 67.5, compass points 6 of 32 and 3 of 16closed · exact · 176
    • F3.4 Three or more stacked tables (Scheidt's patent, "triple matrixes")closed · exact · 177
    • F3.5 The misspelt letters' positions as indices, and their orientationclosed · exact · 178
    • F3.6 The Weltzeituhr's rotating ring as an offset over 26 lettersclosed · exact · 179
    • F3.7 "Point" as a decimal point, a place or a punctuation markclosed · exact · 180
  • ▸5thFamily 2: Masked plaintext, then a keyed stage1 blocked, 7 closed

    • Masks that move or pad letters: reversal, nulls, doublets as nullsclosed · exact · 23 57 75
    • Masks that change letters: affine, Atbashclosed · exact · 74
    • The message written along a grid pathclosed · exact · 92
    • F2.1 Fragments from 0057's letter strike-outs at key lengths 24-25 not yet matched against wordsclosed · exact · 181
    • F2.2 Nulls placed by a ruleclosed · exact · 182
    • F2.3 A plain component with repeated letters: homophones that keep the lengthclosed · exact · 183
    • F2.4 Nulls at both ends of the message (SOE's Dutch section): periodic and progressive keys on the moved known wordsclosed · exact · 23 205
    • A code layer, phonetic spelling, another language: these change the known letters or need a codebookblocked
  • ▸6thFamily 4: Sanborn's visual systems1 blocked, 7 closed

    • Trifid, any cube, periods 2-15closed · exact · 40
    • The plates as a key by position, straight or foldedclosed · exact · 89
    • K4 read from behind the screenclosed · exact · 108
    • F4.1 A route that follows the plates' physical layoutclosed · exact · 184
    • F4.2 Free trifid cubes at periods above 15closed · exact · 185
    • F4.3 The physical anomalies' positions as a keyclosed · exact · 186
    • F4.4 The extra L and HILL as a marker matching YARclosed · by subsumption · 5 17 89
    • F4.5 Antipodes' plate order and spelling differencesblocked
  • ▸7thFamily 10: Alphabets that change after every letter1 blocked, 2 closed

    • F10.1 Chaocipher: two wheels permuted after every letter (1918, disclosed 2010)closed · exact · 187
    • F10.2 Wheatstone cryptograph or Alberti disk advanced by the plaintextclosed · exact · 188
    • F10.3 A key stream from a deck of cards (Solitaire, Mirdek)blocked
  • ▸8thFamily 11: Agency hand systems, now public1 blocked, 4 closed

    • F11.1 Strip ciphers and cylinders with published alphabets (M-94, Jefferson wheel, M-138-A)closed · exact · 189
    • F11.2 Poem code: double transposition keyed by words from a memorized text, then the tableclosed · exact · 190
    • F11.3 The CIA's Polish key box: a key text transposed in a box, then the tableclosed · exact · 191
    • F11.4 Friedman's keyed alphabet box as the key streamclosed · exact · 192
    • F11.5 Poem code with five words picked anywhere in K1-K3, two passes, then the table: 8.1 x 10^17 key pairs, no constraint cuts itblocked
  • ▸9thFamily 8: Hill ciphers1 blocked, 2 closed

    • Hill 2-5, plain and affine, 330 numberings, chained, with a repeating keyclosed · exact · 15 36 44 55 105 116
    • F8.1 Hill with an unknown letter numbering: count the equations firstclosed · exact · 194
    • Settings with too few equations to decide (5x5 affine; long keys with 4x4)blocked · exact · 105 116
  • ▸10thFamily 5: Gromark and number feedback4 closed

    • Two- and three-term rules, bases 2-26 (three-term free alphabet in base 10 only)closed · exact · 13 28 46 54 69 76 107
    • F5.1 Multiplicative feedback; the raised YAR as primerclosed · exact · 196
    • F5.2 Three-term Gromark on a free alphabet in other basesclosed · exact · 195 0262 (the rules left for cost: base 26 three-term lag 6 lag 7 in bases 2-16 two-term lags to 8 base 10 to 10; 2.34 x 10^11 primers nothing fits K4)
    • F5.3 Rules in bases 16-26 on 0054's K1-K3-word alphabet pairs (0054 stopped at 15)closed · exact · 206
  • ▸11thFamily 9: Substitution, then a move1 blocked, 2 closed

    • One-to-one substitution, then any of the 1.86M built routesclosed · exact · 19 106
    • Routes outside every built set: unbounded without a list of hand methodsblocked
    • F9.1 A bounded list of hand transpositions outside the built sets, then one-to-one substitutionclosed · exact · 197 198
  • ▸Family 7: Construction simulation (a check on the others, not a cipher)1 noted, 1 closed

    • K4's two patterns, doublets lining up in sevens and repeats 21 apart, are realnoted · sampled · 56
    • No construction tried makes both patternsclosed · sampled · 18 27 62 65 96
  • ▸Family 13: Ciphertext only: the known words sealed until a reading wins (a way of working across families, issue #85)5 blocked, 1 noted, 10 closed

    • F13.1 Which families 97 letters of English can decide with no known letter: key bits against the judge's reach (reports/f13-solvability-map.md)closed · exact
    • F13.2 The cipher type named from the ciphertext's statistics, power measured first: held-out accuracy 0.28 (chance 0.06), the flat families near chance against random letters; K4 names no family against its shuffles; the one cell under the line (ciphertext autokey) not replicatedclosed · sampled · 231 232
    • F13.3 A stronger English judge: bits per letter it extracts on held-out English: quadgrams 1.75, n-grams without Carter mixed with them 1.95 (189 bits in 97), with Carter 2.38; the judge passes 4-letter soup, not 6-letter soup or word saladclosed · measured · 233
    • ▸F13.4 Ciphertext-only solves in the families F13.1 leaves decidable3 blocked, 1 noted, 6 closed

      • ▸F13.4a Periodic key on the sculpture's 8 table classes, lengths 1-36: K4 inside 200 climbed shuffles (p 0.62); plants found to length 26 (6/8), not at 30-361 closed

        • Lengths 27-36: plants 5/8 at 30, 2/8 at 36; K4 never above all 200 shuffles: sampled at that power (D3)closed · sampled · 234
      • ▸F13.4b Periodic key on 2.08 million dictionary keyword tables, lengths 1-26: K4 inside 100 climbed shuffles (p 0.66); plants found at 5 only (8/8; 4/8 at 10)1 closed

        • Lengths 6-26: the per-column frequency screen loses the true table (plants 0-4/8); K4 never above all 100 shuffles: sampled at that power (D3)closed · sampled · 235
      • ▸F13.4c One free alphabet with a periodic key, lengths 1-19, both forms: plants found to 5 (7/8); K4 inside 50 climbed shuffles there (p 0.37)1 closed

        • Lengths 6-19: plants 2-5/8; K4 inside 100 fresh shuffles climbed the same way (p 0.84), no cell above all 100 (D3)closed · sampled · 236 258
      • ▸F13.4d 27,852 moves before or after the sculpture's tables, lengths 1-20: plants found to 121 noted, 2 closed

        • K4 above its shuffles at table-then-move, key length 2 (p 0.0099, replicated p 0.023 against 300 fresh); reading junk, 3 of 24 known letters; the judge at the edge of chance (14 of 300 at or above)noted · sampled · 237 238 239
        • Lengths 13-20 sampled at power 1-6/8, K4 never above all shuffles (D3); F6.1 (a free alphabet after a move) blocked by 0236's powerclosed · sampled · 237 238 236
        • The replicated excess at B2 (table with a two-letter key, then a move): K4 above 999 of 1,000 fresh shuffles, also without width-19 moves (p 0.004); readings junk, 3 of 24 known letters; F13.4i (0260): planted B2 messages keep their excess under every rotation, K4 under 3 of 96: not a message under this cipher, a property of K4's exact order against the searchclosed · sampled · 237 238 239 259 260
      • ▸F13.4e Hill 2x2 and 3x3, four numberings, every alignment, ciphertext only: 2x2 plants read 8/8, K4 inside 100 shuffles (p 0.73)1 blocked, 1 closed

        • Hill 3x3: plants found 4/8; K4 inside 100 shuffles: sampled at that power (D3)closed · sampled · 240
        • Gromark (cost: 10^5 primers x 364 rules x a free-alphabet climb, about 8 h per text); trifid (0039's annealer fails even with the known letters); M-94 (25! disc orders, beyond 0236's ciphertext-only frontier)blocked
      • ▸F13.4f A puzzle word plus a dictionary word added (0115's space), ciphertext only: 26,040 puzzle-word settings x 2,241,060 dictionary rotations per text, exhaustive per-column screen; plants read exactly 32/32 at 5, 10, 15, 20; K4 inside 100 shuffles (p 0.42)1 blocked

        • Two key-word alphabets plus a key word (F14.1's space): about 4.3 x 10^10 alphabet pairs per text, beyond any screen, and 0235's screen already lost one dictionary table among 2 million beyond length 5blocked
      • F13.4g Chaocipher, two free wheels: given the left wheel the reading is a simple substitution of an index stream, but annealing the left wheel finds wrong wheels above the true one (stream IC 0.070-0.073 against 0.058-0.068, 3 plants); 97 letters do not single out the wheelblocked · desk probe
    • F13.5 Memorable keys from the named sources, scored by English alone: key words, keyword tables, running keys from 36 named texts, autokey primers, keyword columnars; plants found 40/40, K4 inside 100 shuffles (p 0.099)closed · sampled · 243
    • F13.6 A13's structure battery on every reading, not only 24-of-24 ones: the readings are chosen by English, so masks that break English never reach it (every nth 0/4, affine 0/4, columns 1/4) and those that keep it are already found (X every kth, nulls 4/4)blocked · desk probe
    • F13.7 The object: the copper's letters, the ?, the line breaks, Antipodes' K4; needs the photographs (#81)blocked
  • ▸Family 14: Scheidt's mask, keyed by words: the blocked free-alphabet cells, shrunk by his "key word/s" (issue #93)1 blocked, 7 closed

    • ▸F14.1 Two dictionary key-word alphabets plus a key word: all 4.3 x 10^10 pairs per form matched exactly, key lengths 1-24; no key word on K4 or any shuffle, pairs only where the shuffles have them; plants 12/12; K4's count at 16 replicated, p 0.009, not a flag1 blocked

      • Key lengths 25-26: no dictionary word that long; 10^6-10^9 pairs fit by chanceblocked
    • F14.2 A remembered route plus key-word alphabets in F6.1 and F6.2's cellsclosed for 96.9% of settings · exact · 0271: no key word on K4 or any of 50 shuffles plants 12/12; 41 986 settings unverified by a listing flaw open for F14.2a
    • F14.3 Duress reading: a running key with the message and the key both English; 0125's 3.7 million forced keys ranked by MIX8m, read by the judge: plants 20/20, K4 inside 50 shuffles (p 0.59)closed · sampled · 246
    • F14.4 How firm is letter for letter (A2)? The 2010 Times wording ties carved 64-69 to BERLIN, three dinner accounts say one-to-one, the 2019 retreat refuses more without contradicting; A2 stands, length-changing digit systems stay outclosed · by reading
    • F14.5 Scheidt's '98 characters': F14.1 with one carved letter lost, the 23 places that change anything: no key word on K4 in any; plants 23/23closed · exact · 247
    • ▸F14.6 Mix and match, up to three steps from K1-K3's operations: 11 canonical forms, 7 closed and cited, 4 run (reports/f14.6-grammar.md)3 closed

      • F14.6a Three puzzle words summed; a puzzle keyword table with a puzzle word plus a dictionary word: nothing fits K4 or its shuffles; plants 16/16closed · exact · 248
      • F14.6b Puzzle keyword alphabets around a move, key word of any length: 2.47 x 10^9 settings, exact; no key word on K4 or any shuffle; plants 8/8closed · exact · 249
      • F14.6c A move, then a puzzle word plus a dictionary word summed: 1.45 x 10^9 settings, exact; no word on K4 or any shuffle; plants 8/8closed · exact · 250
    • F14.7 Plaintext style as the mask, on the judge bench: the judge sees telegraphese (1.00) and abbreviated (0.97) English, not vowel-dropped (0.13); MIX8m and quadgrams see noneclosed · measured · 251
  • ▸Family 15: K1-K3 as a guide to K4's content: the plaintext is Sanborn's own prose (#117), so model what it says, then use that as extra known plaintext (issue #118)3 closed

    • F15.1 The content model, by reading: K1-K3's themes, voice and mechanics, every statement on K4's content, the layout the 24 letters fix; written as content hypotheses with predicted wordsclosed · by reading · `reports/f15.1-content-model.md`: seven hypotheses H1-H7
    • F15.2 Drafts: hundreds of 97-letter texts in Sanborn's voice, one set per content hypothesis, the 24 letters in place, plus a held-out setclosed · 329 corpus drafts (Opus) + 28 held-out (Sonnet, another prompt) in `data/drafts/drafts.jsonl`
    • F15.3 Power first: on planted drafts, do the other drafts help a search find the plant, beyond the 24 letters alone?closed · measured · 0268: English 18/24 drafts prior 16/24 (bar +4 not met)
    • F15.4 Probable phrases at fixed places: phrases the drafts agree on, and K1-K3's own phrases ("comes back to the first part"), as extra cribs in the cells blocked for lack of checksclosed as weak by F15.3's rule (0268)
    • F15.5 Each draft as a full crib with slip tolerance: the forced key read for structure, against copiesclosed as weak by F15.3's rule (0268)
    • F15.6 Past readings rescored by closeness to the drafts' content, the judge kept blindclosed as weak by F15.3's rule (0268)