TanachCodes.com
Search Tanach (Hebrew Bible) for Equidistant Letter Sequences (ELS) — Torah Codes — and many other letter-pattern modalities, purely in your browser.
Enter a Hebrew word or phrase of at least 4 letters, pick a search modality, and the app scans the chosen scope for every match.
- 100% client-side — no servers, no API calls after load.
- Offline-capable — runs entirely from your browser cache after the first visit.
- Source text — Westminster Leningrad Codex (WLC), public domain.
View source on GitHub
Search modalities
Plain text
חיפוש טקסט
(chipush tekst)
Exact consonantal substring search. The text is nikud-stripped, ta'amim-stripped, and sofit-folded, so the input phrase is matched as a literal run of letters.
When to use: when you know the literal phrase you are looking for.
Example: בראשית → matches at the start of Genesis 1:1.
Equidistant Letter Sequence (ELS)
דילוגי אותיות שוות
(dilugei otiyot shavot)
The classical Torah-codes algorithm. Picks every Nth letter (for every N in your skip range) and looks for the search phrase in the resulting stream. Forward, reverse, or both directions.
When to use: to look for a phrase encoded at a fixed letter-skip — the original Witztum/Rips/Weissmandl convention.
Example: תורה with skip 50, forward → finds it spelled at every 50th letter starting near the beginning of Genesis (the classical Weissmandl finding).
ELS — Atbash variant
דילוגים באתב״ש
(dilugim ba-atbash)
Same as ELS, but the corpus is first run through the atbash cipher (א↔ת, ב↔ש, …) before searching. Toggle the Cipher field in the ELS parameters panel.
When to use: to look for a phrase encoded at a fixed skip under atbash substitution.
Example: יהוה with cipher = atbash → searches for the atbash-encoded form of the divine name across all skips.
Word position / Notarikon
מיקום אות במילה / נוטריקון
(mikum ot ba-mila / notarikon)
Picks one letter from each word at a fixed structural position (first, second, third, middle, second-to-last, or last) and searches that derived stream for the phrase. With first or last, this is the classical notarikon of rashei tevot / sofei tevot.
When to use: to find phrases hidden in the initials, finals, or any per-word fixed position.
Example: letter-position = first, every word, phrase = אדם → finds passages whose consecutive word-initials spell א-ד-ם.
Gematria
גימטריה
(gematria)
Compute the gematria of your search phrase under one of four schemes, then find every contiguous run of 1–3 words (default) whose letter-value sum equals it.
- Standard — מספר הכרחי (mispar hechrachi). א=1, …, י=10, כ=20, …, ת=400.
- Ordinal — מספר סידורי (mispar siduri). 1..22.
- Reduced — מספר קטן (mispar katan). Standard mod 9 (0→9).
- Atbash — גימטריית אתב״ש (gematriyat atbash). Standard value of the atbash-substituted letter.
When to use: to explore numerical equivalences between phrases.
Example: אהבה with scheme = standard → target 13; matches include אחד and other 1–3-word spans summing to 13.
Cluster ELS
צירוף דילוגים
(tziruf dilugim)
Run two ELS searches, one per phrase, then emit pair matches whose hits land close to each other in the 2-D wrap of the text at the smaller of the two skips. Each pair shows as two linked rows.
When to use: to look for spatially related code-pairs — the rigorous half of the original Witztum-Rips claim.
Example: phrase A = יהוה, phrase B = אלהים, max table area = 20 → finds pair occurrences where both phrases sit inside a small bounding rectangle when the text is wrapped at the smaller skip.
Anagram
אנגרמה
(anagrama)
Slide an N-letter window across the corpus and emit every position whose 22-letter histogram exactly matches the search phrase's. Optional "word-aligned only" toggle restricts hits to spans starting at a word boundary.
When to use: to find permutations of a word's letters embedded in the consonantal text.
Example: שלום → finds 4-letter spans whose letters are any permutation of ש-ל-ו-ם.
Proximity / co-occurrence
קרבת מילים
(kirvat milim)
Take two phrases and find every place where both occur within the same window — either "same verse" or "within N letters".
When to use: to find verses or short letter ranges where two phrases co-occur.
Example: phrase A = אהבה, phrase B = שלום, window = same verse → returns every verse containing both words as plain text.
Deep Search
חיפוש עמוק
(chipush amok)
Runs all single-phrase modalities in sequence on the same input phrase: ELS (×2 ciphers), plain text, gematria (×4 schemes), anagram (×2 alignments), and the full 6×5 word-position grid. Total = 39 sub-searches. Cluster and proximity are not included — they require a second phrase.
When to use: when you want one combined hit list across every relevant modality without running them one at a time.
Example: Deep-search תורה → a single ranked result list across all 39 sub-runs.
Results filters
- Book filter — narrow to one book.
- Min skip per book only — for ELS, hide all but the smallest-skip hit in each book.
- CSV export — download the visible result list.
Search history is persisted in your browser; click the history button to replay any past search.
Torah Codes: A Brief Research Summary
The idea that the Torah hides messages in equally-spaced letters predates the computer-era controversy around it. The 13th-century Spanish commentator Bachya ben Asher (Rabbeinu Behaye) gave a four-letter ELS example tied to the zero-point of the Hebrew calendar, and similar ideas appear in Moshe Cordovero's Pardes Rimonim (16th c.) and in writings attributed to the Vilna Gaon. The modern story really begins with Rabbi Chaim Michael Dov Weissmandl, a Slovak Talmudist and mathematician who, before WWII, claimed to have found meaningful words at fixed letter intervals in Genesis by counting letters on paper. His findings were published posthumously in Toras Chemed (1958), and his best-known example — the word תורה occurring at every 50th letter from the start of Genesis — became the touchstone for everything that followed.
The technique became a public scientific question with Witztum, Rips and Rosenberg's 1994 paper in Statistical Science, the so-called "great rabbis experiment." They reported that ELSs of the names of 32 prominent post-biblical rabbis cluster suspiciously close to ELSs of the day-and-month of those rabbis' births and deaths, with a randomization test giving a p-value of about 0.00002. The journal's editor Robert Kass presented the article as a "challenging puzzle" rather than an endorsement, and a follow-up experiment by NSA cryptanalyst Harold Gans, using cities of birth and death, was put forward as independent confirmation.
The rebuttal came in the same journal five years later: Solving the Bible Code Puzzle by Brendan McKay, Dror Bar-Natan, Maya Bar-Hillel and Gil Kalai (1999), reviewed by four senior anonymous statisticians. Their case was that medieval Hebrew rabbis have many valid name spellings and many honorific appellations, and that WRR's specific list — which actually determines the outcome of the experiment — could be lightly perturbed within those rules to make Genesis look ordinary while making the Hebrew translation of War and Peace score just as strongly. Editor Kass wrote in his introduction to the rebuttal that the puzzle had been solved. Separately, in response to Michael Drosnin's bestseller The Bible Code, McKay located apparent ELS "predictions" of the Kennedy, MLK, Rabin and Indira Gandhi assassinations in Moby-Dick, demonstrating how much you can find by eye in any long text once spelling and search parameters are flexible.
The mainstream statistical view today is that the original WRR result is an artifact of researcher choices in name selection rather than evidence of an embedded code, and even Witztum and Rips have publicly distanced themselves from Drosnin's predictive claims. Within Orthodox Judaism the picture is more mixed. Some teachers (see Aish.com's overview) continue to cite the codes as a kind of mathematical apologetic, often invoking older statements of Nachmanides, the Vilna Gaon and the Maharal about "hidden" content in the Torah. The older sources are mostly about classical exegetical methods such as gematria, notarikon and temurah applied within the surface text, and most traditional commentators read those claims as homiletical rather than as a predictive ELS framework. McKay's own summary on his Moby-Dick page captures the academic side fairly bluntly: the patterns violate no laws of probability, and once you understand the method you can produce similar matches in any sufficiently long text.