[bxguff]

the LLM ouroboros starts with models being used to create training data for models.

[jstarfish]

I'm actually looking forward to this because the result is going to be hilarious-- the culmination of literally every slippery-slope argument ever as the models reinforce their own biases over time.