dear GPT-5, i’m pretty sure “storagely” is not a word, but that oddly makes a...

dear GPT-5, i’m pretty sure “storagely” is not a word, but that oddly makes a ton of sense so maybe i’ll start using it

honestly this model really does feel “big”, but in a totally different way than R1, K2 & Opus. like a whole ass new direction for depth

it’s like that nerdy kid whose words can’t keep up with their brain

my reading on softmax & entropy in LLMs has led me into economics

you can indeed calculate the “temperature” (same thing as the LLM parameter) of wealth redistribution across the years for an economy

and it’s been done:

apps3.aps.org/aps/units/ma...

expanding slightly — a low temperature drives “wealth” (or logit values, for LLMs) into fewer hands

whereas higher temperatures will redistribute equally without regard to the model’s (economy’s) biases

so temperature is a parameter that can exaggerate the internal bias or structure