LANDMARK · 5 MIN · USE
Iterate deliberately
After this landmark, you can improve an AI output by describing the exact change you want instead of asking vaguely or regenerating at random.
When an output is close but not right, there are two ways forward. You can reroll — regenerate and hope the next draft is better — or you can name the specific gap and ask for that change. Rerolling is a lottery; it can lose the good parts of the draft you already had. Deliberate iteration keeps what worked and fixes what didn’t, because you told the model precisely what “didn’t” means.
This is the canonical concept. It stays the same across learner lenses so personalization never changes the underlying facts.
◌ EVERYDAY LENS
What this looks like for you
Don’t say “make it better.” Say what’s off: “too formal,” “cut it to three sentences,” “lead with the cost.” Specific feedback gets a specific fix.
MAKE A DECISION
An AI drafts a solid email but the opening is too long-winded. You want it tightened. What gets you there fastest?
CARRY THISNext time an output is close, resist the reroll. Write one sentence naming exactly what’s wrong, and edit from there.