Idea for thinking about [[Natura Naturans]], Makina Naturans thinking about [[machina Naturans]] we might think that it's quite easy to give LLMs to get them to do just a completely random weird free associative output in fact maybe that's what post-training is rectifying we need to look into this