First words

A sample counts as FIRST WORDS when it contains a run of 6 or more tokens in a row that are all real words, and:

  • real words means the SCOWL en_US size-50 word list plus a fixed crypto-slang list (gm, ser, wagmi, lfg and about 230 more);
  • every neighbouring pair in the run also appears next to each other at least 3 times in a 14 MB public-domain English reference (pairs with a slang word are exempt), so word salad like "wall wall moved moved" does not count;
  • at least 3 different words, of which at least 2 are not function words ("the of and to" does not count);
  • no 6-word window of the run appears in the coin's own corpus, so memorised text never counts;
  • the sample passes the slur and scam filter twice.

The first qualifying sample is the coin's FIRST WORDS. It pins the coin to the top of the nursery and gets a share card and a 10 s clip. If a holder's submission shares 2 or more 4-word sequences with the sentence, the card says "learned this from" with that wallet shortened.

ABOUT TO SPEAK shows when a coin's validation bits per byte is at or below the calibrated threshold (the median bits per byte at which the lab runs' longest real run first reached 5 words), or when its latest sample has a 5-word real run.