Full threadteekert·Probably this was already done at Google, Meta, X and OpenAI, before training their LLMs.View on HN