Workbench
Configure a transposition + substitution pipeline and score it against K4’s 24 known crib positions. All computation runs locally in your browser (no data is sent to any server).
Your odds of solving K4 right here (spoiler: astronomical)
This workbench exposes roughly 1035 possible configurations. At 3 per minute, 8 hours a day, 365 days a year, you would finish in about 1028 years (roughly a quintillion universe lifetimes). Bring snacks.
You are 1012× more likely to win the Powerball twice in a row than to stumble on the answer here. And yet, someone, someday, might try an “obvious” keyword that over 35 years of PhDs missed and crack the last unsolved message at CIA headquarters. That’s cryptanalysis.
Over 671.0B+ configs have been evaluated by this project. Most of the remaining space is not honestly enumerable.
What we know about K4 (research context)
Sanborn stated K4 uses two systems of encipherment. Pure transposition is impossible (CT has 2 E’s, cribs need 3), so at least one layer is substitution. Beyond that, the architecture is open: the live questions are about ordering, segmentation, and whether any procedural layer is present.
Eliminated proven impossible
- All single-layer substitution on the raw 97 characters (periodic Vigenère/Beaufort/Variant Beaufort, periods 1–26, both AZ and KA alphabets)
- Pure transposition (CT has 2 E’s, cribs need 3)
- Columnar transposition × periodic substitution (47M+ configs tested)
- Null mask + periodic substitution on any 73-char extract (periods 1–23)
- Autokey on raw 97 (exhaustive single-letter + 1M dictionary keys)
Open territory viable
- Non-standard transpositions (serpentine, spiral, Myszkowski) + substitution
- Autokey / running key after undoing transposition
- Custom or Quagmire tableaux
- 73-char hypothesis with non-periodic cipher after null removal
- W-delimiter hypothesis (2026-04-17): the 5 W characters at positions 20, 36, 48, 58, 74 divide K4 into 6 segments of 20|15|11|9|15|22 chars. Removing them eliminates the width-21 vertical bigram anomaly entirely, making them a live structural hypothesis surface. Use the “W-segment” transposition and “W positions only” null mask to explore this.
Watch for 23/24
Some researchers explore whether K4 could use a delimiter convention analogous to
marker-like characters elsewhere in Kryptos. That is a hypothesis, not a fact. If
you hit 23/24, inspect whether a single site behaves like a marker rather than
assuming the miss is random. One concrete variant: carved W may decrypt
to a rare plaintext letter such as X, Q, or Z
serving a punctuation-like role.
Quick presets
Vigenère + KRYPTOS
K1–K3 method eliminated
Beaufort + KA alphabet
Kryptos-keyed Beaufort eliminated
Columnar w=7 + Vig
Two-system model open
Rail fence + Beaufort
Non-standard transposition open
Serpentine + Vig
Boustrophedon reading open
DEFECTOR 15/24
Historic high score disproven
W-delimiter (92-char)
Single-layer saturated multi-layer only
Try your own
Blank slate
K4 Ciphertext
73-char hypothesis: 24 of 97 are nulls.
Extracted CT (0 chars, 0 nulls removed):
6 W-delimited segments as grid rows: 20 | 15 | 11 | 9 | 15 | 22 chars (sum = 92). Set null mask to “W positions only” first to extract the 92-char text.
Enter a substitution key above to see results.
Transposed CT
Plaintext
Derived keystream at crib positions
Bean constraint (k[27]=k[65]): -- (PASS requires the correct solution; always FAIL on raw 97).
Period consistency analysis
For each period, counts how many crib-position keystream values conflict (share a residue class but have different key values). Zero conflicts = consistent with that period.
Session history (0 attempts)
Cipher toolkit inspired by enigmator by Merricx.