Amateur linguist claims AI-built tooling helped crack Minoan Linear A

Tom Di Mino, a self-taught AI engineer and amateur linguist based in the Hudson Valley, says he has deciphered Linear A — the Bronze-age Minoan script that has stumped scholars for more than a century. His claim, now under review by linguists at Rutgers and Cambridge, is that the script encodes an extinct Semitic language ancestral to Hebrew, Aramaic, and Arabic. The Semitic angle isn’t new (Cyrus Gordon argued it in 1957), but earlier attempts never yielded working translations or won acceptance. If it holds up, it would rank alongside the 1952 decipherment of the related Linear B script as a landmark in the field.

The technical hook for an engineering audience is how he got there. Linear A shares about 60 syllabic signs with the already-decoded Linear B, so the sounds of overlapping symbols were known but their meaning wasn’t, and 13 Linear A-only signs had no accepted phonetic values at all. Di Mino’s breakthrough came from a formulaic set of peak-sanctuary prayer inscriptions where every word but the opening verb was already readable. Mapping the unknown verb to the Semitic tri-consonantal root N-W-Y (‘to dwell’) let him crack the formula and, he argues, cascade outward to logograms and even to disputed Linear B readings.

Critically, Di Mino built the analysis pipeline using Claude Code, generating Python scripts to query and cross-reference the digitized GORILA and SigLA corpora and run systematic hypothesis testing at a scale impractical by hand. The output so far includes proposed readings for 40 signs (including the 13 previously unknown ones), a 408-term lexicon, and a draft manuscript intended for peer review. The result is unverified and the author discloses a personal acquaintance with Di Mino, so the appropriate stance is cautious interest — but it’s a striking example of LLM-assisted tooling accelerating brute-force scholarly work.