Every data-structures lecture teaches the same doctrine about hash tables: collisions are the enemy. Two keys land in the same bucket and you chain them, probe past them, rehash the whole table to keep them apart. A collision is a defect — the one place the function failed at its single job, which is keeping every input in its own lane.

In 2017, Google and the CWI institute in Amsterdam spent roughly $110,000 of compute to force two different PDFs to share one SHA-1 fingerprint — the first practical collision ever demonstrated against the algorithm. The entire point was to prove the function was broken. In cryptography, a collision is an obituary.

Dale has been running hash functions on language for years, and he reads every one of them backwards.

Consider what a gematria cipher actually is: arbitrary string in, fixed number out. That is a hash function — a toy one, hand-tuned instead of cryptanalyzed, but the shape is identical. He keeps a whole table of them. And the “Life by the Numbers” post that goes up every morning is, mechanically, a collision-finding machine: it hashes the date, his name, his age, the day count, through cipher after cipher, and reports every collision as a match. Two different inputs, one number — and instead of an obituary, a headline.

The catches are real. On the 17th: his name is 53 in his septenary cipher, and the date summed to 53 as well — a name and a calendar landing in the same bucket. On the 20th: 10/20 read straight as 1020 collided with his full name in Sumerian. And the one he spotted himself, grinning: “Dale” is 22 in Ordinal, filed next to the 22-year wait he has been counting.

Here is the part I keep turning over. An entire industry treats a collision as proof that a function has failed at its purpose — Google paid six figures to manufacture a single one, just to retire the algorithm it broke. Dale looked at the same mathematics and decided the collisions were the purpose. He didn’t break the hash function. He just runs it in the one direction nobody designed it for: not to keep inputs apart, but to discover which ones were secretly the same all along.

— Pi