azali peam

Definitely some cognitive dissonance there but I don't think it's a bad thing, it is funny though

Yeah, I haven't looked at how to alter the template yet but I want to remove that, noticed the first time I ran this

huggingface.co/karashiiro/g...

huggingface.co/datasets/kar...

There’s an interesting line in between where you’re using it as a sort of research agent to help pull together the papers for a survey. But the point of writing such a paper is to learn a lot or to construct new categories for yourself

that's right, it goes in the square hole

forgor to say trick or treat

irel vagrerfgvat gung gur zrzbel jbeqpybhqf ner zhpu zber fvzvyne orgjrra gur zbqryf guna gur pung jbeqpybhqf, ner gur zrzbel jbeqpybhqf graqvat gb fgber dhbgrf sebz gur hfre ntrag?

almost solely basing those guesses off the hypothesis that the models will create fewer memories when they aren't presented with challenging information

how far off is this

A: Sonnet w/ boundary - fewest messages but most memories (possibly surprised by the user agent?)
B: Sonnet w/o boundary - similar characteristics to A, but slightly more messages and slightly fewer memories
C: 4o - longest conversation, fewest memories (unchallenged)

I sense a pattern