There's a new large language model called Talkie, trained only on English text from before 1931. It can write in a charming old style, and then say something obviously false or downright mean. Looking at those responses felt like sitting on the meditation cushion. It got me thinking about when and how I have stopped updating.

