How-to · INT · INT.2

Say what NOT to do — negative constraints work

Explicit must-nots (no hype words, no invented stats, no preamble) shape output as much as positive instructions.

Open any chat tool and paste: 'Write a 3-sentence product update. Must NOT include: exclamation points, the words revolutionary/seamless/game-changing, or any invented metric.' You're done when the output violates none of the three bans — or you caught the violation and called it out.

Worked example

Everyone tells the model what to do. Pros also tell it what not to do: no invented numbers, no buzzwords, no apologies, no preamble. Negative constraints are how you kill the tells that make drafts sound machine-made.

← All How-tos

Say what NOT to do — negative constraints work · ConeheadAI How-to | ConeheadAI