About this project
What "Bible codes" are, where the claim came from, how these tools test it, and how the whole site is built.
This site is an honest, interactive exploration of the "Bible code" idea: the claim that meaningful words are hidden in the Hebrew Torah as equidistant letter sequences — words spelled out by reading every d-th letter.
The whole debate turns on a single question: do these patterns appear in the Torah more often than chance predicts? That question can't be answered by looking at the Torah alone — any striking-looking pattern needs a baseline. So every search on this site runs simultaneously on the real Torah and on four comparison texts of identical length that were built to contain no hidden design. If a pattern is special to the Torah, it will stand out against them. If similar patterns arise in all five texts at similar rates, that is what ordinary combinatorics predicts for any long text over a 22-letter alphabet.
The stance here is deliberately neutral. The history is presented respectfully, the searches run live and unfiltered, and the numbers are whatever they are — you can judge the evidence yourself. Nothing is precomputed for you to take on faith; the tools re-derive every result in your browser as you watch.
Three ideas are enough to understand both the claim and the catch.
Strip a Hebrew text of spaces and vowels and it becomes one long chain of letters. An equidistant letter sequence (ELS) is a word spelled out by starting anywhere in that chain and jumping a fixed number of letters at a time — the skip — forwards or backwards. Here are the first sixteen letters of Genesis; start at the third and take every fourth:
בראשית ברא אלהים… — letters 3, 7, and 11 spell אבל ("but", or "mourning"). A real sequence, at skip 4, three letters into the Torah.
The dramatic letter-grids in Bible-code books are made with one move: write the same chain in rows of a chosen width. If the row width equals a word's skip, that word stands up as a vertical column; nearby widths turn it into diagonals. The text never changes — only the wrapping does. Try it on the same sequence:
This is worth sitting with: the same three letters, at the same fixed skip, look like a striking column at one width, a diagonal at another, and nothing at all at a third. The sequence doesn't move — the picture does. Every published code table is a choice of wrapping that makes one particular finding look geometric.
Whether such sequences are surprising is a question of arithmetic, not mystery. A search for one short word across the Torah examines an enormous number of starting points and skips:
With a 22-letter alphabet, the expected number of chance appearances of a 4–6 letter word across that many trials ranges from dozens to thousands — in any text of this length, sacred or scrambled. Allow several related words, several spellings of each, and any window of text for them to cluster in, and striking-looking coincidences stop being surprising and become a statistical guarantee.
Rabbi Michael Dov Weissmandl noticed letter-skip patterns in the Torah in the 1940s, counting by hand. The modern story begins in 1994, when Doron Witztum, Eliyahu Rips and Yoav Rosenberg published "Equidistant Letter Sequences in the Book of Genesis" in the peer-reviewed journal Statistical Science. They reported that the names of famous rabbis appeared as ELSs in Genesis close to their dates of birth and death, at a proximity they calculated was far beyond chance.
The idea reached the public through Michael Drosnin's The Bible Code (1997), a bestseller that used it to claim hidden predictions of assassinations and disasters — a popularization that most researchers on both sides considered unfounded.
In 1999, Brendan McKay, Dror Bar-Natan, Maya Bar-Hillel and Gil Kalai published "Solving the Bible Code Puzzle" in the same journal. They argued that the 1994 result depended on flexible choices made while preparing the experiment — which spellings of each name, which appellations, which date formats — and showed that with equally defensible choices the effect vanished. Crucially, applying the same methods to a Hebrew translation of War and Peace produced "codes" just as impressive as those in Genesis. The tools on this site let you re-run that comparison yourself.
Every text is exactly 305,641 letters — the Torah's consonant count — with final letter forms collapsed to their base form (ך→כ and so on), the standard convention in this field. Four of the five are controls that encode nothing:
The two "random" texts test raw combinatorics; the word-salad and Wikipedia texts test whether real Hebrew vocabulary alone is enough to manufacture codes. McKay's War and Peace control was of this last kind — real literature, not scripture.
An ELS is a word spelled by every d-th letter of the text, forwards or backwards. Searches here cover skips from 2 up to your chosen limit, in both directions. Skip 1 — the plain, unskipped text — is excluded, and so are words shorter than four letters, because those turn up constantly at some skip in any text and carry no information.
Because Hebrew words are often spelled more than one way, searches optionally include plene/defective variants (dropping internal vav/yod) — the same latitude code proponents use. A cluster is the window of at most ~1,500 letters that contains the most distinct search words, each counted once, with ties broken toward the tighter window. The display grids wrap the text at one cluster hit's actual skip, which is what makes the sequences line up as vertical or diagonal streaks.
Where a search reports an expected-by-chance count, it assumes independent letters drawn at that text's own letter frequencies — a deliberately simple null that ignores real spelling correlations, so it is a floor, not the last word. The forthcoming per-search significance test (a shuffle-based empirical p-value) will address that directly.
The Torah text is the Hebrew consonantal text from Sefaria. The Wikipedia control and the word-salad sample are drawn from Hebrew Wikipedia. The Gematria calculator's ~101,000-word list combines the most frequent words in a large Hebrew Wikipedia sample with common dictionary words (Hspell, intersected with the CC-100 web corpus) so everyday vocabulary isn't missed. Random texts use fixed seeds so they are reproducible. Hebrew renders everywhere via an embedded Noto Sans Hebrew subset, so no system Hebrew font is required.
Self-contained static pages. Every text and word list is baked in at build time, and all searching happens in your browser — there is no server, no database, no analytics, and no cookies. Nothing you type is uploaded or stored anywhere but your own machine. The whole site is designed to cost nothing to run at any level of traffic, and its source will be released as open source.