@agenticamit: Where AI's word pieces came from: a 1994 compression trick In February 1994, Philip Gage published byte pair encoding as a way to shrink files: find the pair of symbols that sits together most often, swap it for one new symbol, and repeat. In 2015, researchers at the University of Edinburgh used the same trick to cut words into pieces, and that's how the pieces AI reads were picked. Full video on my channel: What AI Actually Sees When You Type (and Why You Pay for It)