Linear B · deciphered 1952 structure → sound
Artificial Wasteland  ·  Language  ·  decipherment

The Grid That Spoke Greek

A Bronze Age script was cracked without anyone knowing what a single sign said. First the signs were sorted into a grid by pure structure — which share a vowel, which share a consonant — with no sounds at all. Then four guessed place-names pinned a handful of rows and columns, and signs that appeared in no place-name came out forced — and what they spelled was Greek. One of those forced words was tripod. Then a tablet no one could yet read turned up with that very word, beside a three-legged cauldron.

I · structure before sound

How to link two signs without knowing either

II · the machine

The grid that spoke Greek

Here is a corner of the grid — a column for each vowel, a row for each consonant. Each cell holds one sign, shown first only by its catalogue number. No sounds are assigned. Walk it forward.

from the four place-names forced by the grid — in no place-name

III · the test it could not rig

The pictures below are schematic — the tablet’s own ideograms show the same leg- and handle-counts beside each word.

IV · what the grid cannot do

Where the method runs out

    colophon · the check
    ◆ dataset verification: loading…

    Not “AI art.” This is standard, textbook history of the decipherment, made operable. Nothing here is an original finding; the only new thing is the form — the grid run as a machine. The interactive is a faithful subset of an ~87-sign syllabary; the propagation it shows is true to the method, smaller than the historical one.

    Sources.