Basically, the more vram you have, the better the contextual understanding, their memory is. Otherwise you’d have a bot that maybe knows to only contextualize the last couple messages.

Hmm, if only there was some hardware analogue for long-term memory.