The crossword looks like a 20th-century invention: Arthur Wynne printed the first “word-cross” puzzle in a New York newspaper in December 1913. But the idea itself — laying text into a grid so that it reads in several directions at once — is a good thousand years older. Classical Sanskrit poetry has a whole genre of citrakāvya (“figure poetry”), and within it a family of bandha (“binding”) forms: a verse is written out as a figure — a square, a wheel, a lotus, a zigzag — and the figure dictates how it is read.
The form closest to the crossword is the sarvatobhadra (“auspicious from every side”). The thirty-two aksharas of a verse are laid out in an 8×8 square so that every row and every column reads as a verse pāda — left to right, right to left, downwards and upwards. Bhāravi (6th c.) has such a verse in canto 15 of the Kirātārjunīya (15.25, devākānini kāvāde…); Māgha (8th c.) has his in canto 19 of the Śiśupālavadha (19.27). This is the crossword intuition in its purest form: one cell — one akshara, and every cell works in several readings at once. The Western parallel is the Latin Sator square (SATOR AREPO TENET OPERA ROTAS), attested since the 1st century — also a square readable in four directions, but made of five words, whereas the sarvatobhadra carries a full verse with grammar and meaning.
Three things are worth seeing at once. First, the unit of a bandha is not the letter but the akshara, the Devanagari syllable; our akshara crossword is built the same way, and that is no coincidence — a syllable cell is more natural for Sanskrit than a letter cell1. Second, in a crossword the crossings are rare points where two words share a letter; in a bandha every cell belongs to several readings simultaneously: a sarvatobhadra “crosses” in all 64 cells. Third, a crossword is assembled from independent words, while a bandha is one continuous verse that the poet managed to fold into rigid geometry without losing metre or sense. This is why citrakāvya was called a hard game: Daṇḍin’s Kāvyādarśa files such figures under duṣkara — “the difficult to perform”.
Sanskrit theorists themselves argued about the price of this virtuosity. Ānandavardhana in the Dhvanyāloka (9th c.) ranked citra as the lowest grade of poetry: where all attention goes to form, the “resonance” of meaning has nowhere to arise. Yet poets kept writing bandhas — precisely because they were an examination in absolute command of the language. Canto 19 of the Śiśupālavadha is such an examination, over a hundred stanzas long: next to the wheel and the square stands, for instance, stanza 19.3, each quarter of which uses a single consonant only (jajaujojājijijjājī… — only “j”, then only “t”, “bh”, “r”), while Bhāravi has a stanza built on the single consonant “n” (Kirātārjunīya 15.14, na nonanunno…). Tradition pits the two poets against each other for a reason: the Kirātārjunīya set the figure patterns, and the Śiśupālavadha answered each one — and added its own.
It is telling where the bandha lived on: it “signed” poems (as Māgha signed his wheel), it was demanded of court poets as a certification, it was taught — Rudraṭa in the Kāvyālaṃkāra (9th c.) was the first to systematize the forms and advised weaving the author’s name into them. Squares and wheels were carved on temple walls: a figure verse worked as text, as ornament and as amulet at once — “sarvatobhadra” literally means “auspicious from every side”.
Sarvatobhadra — the 8×8 square, “auspicious from every side”: rows and columns yield the verse pādas in both directions (Śiśupālavadha 19.27).
Cakrabandha — the wheel: three lines of the verse lie on three diameters with the shared hub “ra”, the fourth runs around the rim, and aksharas 3 and 17 of each diameter spell the hidden signature mā-gha-kā-vya(ḥ)-mi-daṃ — “this is Māgha’s poem” (19.120). A medieval poet “signed” his verse inside its own geometry — a millennium before watermarks.
Gomūtrikā — the “cow trail”: two lines whose even-position aksharas coincide, so they can be read both straight and in a zigzag (19.46).
Padmabandha — an eight-petalled lotus with a shared pericarp; we found no verifiable classical specimen with a published reading scheme in open sources, so that page carries an honestly labeled new teaching composition built by the classical rule.
Gomūtrikā — solve it yourself — a teaching composition from first-year vocabulary: type two sentences into the 2×8 grid and watch the cow-trail come alive. Every page in this section now has a fill-in mode: hide the aksharas and rebuild the verse cell by cell, with checking and reveal.
Each page carries a real verse from the Śiśupālavadha (GRETIL e-text, Göttingen) laid out in its figure. Tap a reading path — the cells light up in reading order, and the panel shows the path text in IAST with a translation. Every path is checked mechanically: the build script verifies that concatenating the aksharas along each path reproduces the verse exactly. The reading schemes are checked against Shila Chakraborty’s 2018 study of the Śiśupālavadha; where tradition allows a licence (the visarga in Māgha’s signature), we say so right on the page.
The bandha is not a museum curiosity. It is the most vivid demonstration that the akshara is the real unit of Sanskrit writing: poets measured their grids in syllables, not letters. Half an hour with a sarvatobhadra builds the habit of reading Devanagari in any direction and tracking aksharas as cells — the same skill our crossword trains. And it is simply beautiful: a verse that stays a verse no matter how you read it.
1 Indian syllable division does not coincide with English or Russian syllabification. An akshara is a unit of Devanagari writing, not of hearing: a vowel on its own, or a consonant — including a cluster of two or three consonants written as a single ligature — followed by a vowel, so an akshara is always “open”, ending in a vowel (possibly with anusvāra or visarga): ga-ccha-ti, kṣe-tra. English and Russian syllabification is phonological and freely closes a syllable with a consonant (cross-word, пас-тух); Devanagari instead pulls every pre-vocalic consonant into the following akshara, while a final consonant with no vowel (marked with virāma, like the m of śubhām) attaches to the preceding one. Bandha grids and Sanskrit metrics are measured in aksharas.
An 8×8 square: every row and column is a pāda of the verse, in both directions. Śiśupālavadha 19.27.
A wheel with three diameters, the shared hub “ra” and Māgha’s hidden signature. Śiśupālavadha 19.120.
Two lines read straight and in the “cow-trail” zigzag. Śiśupālavadha 19.46.
An eight-petalled lotus with the shared pericarp “ma” — a new teaching composition.
A solvable gomūtrikā from first-year vocabulary: type two sentences so the cow-trail works. Fill-in mode with per-akshara checking.