MrCheeze · May 15

this one too actually

tachikoma · May 15

but i think these images are being converted/compressed into token space like text, not being retained in full visual fidelity in the context window. that's why there's forgetting, unlike however human minds represent reality in some richer format.