One day I want to assemble a corpus of every single thing I've written online over a roughly 20-year span. That'll be tough, since a lot's been either linkrotted away or behind paywalls. In the meantime, trying to see if I can finetune an open model to match my voice (Joe Weisenthal)