Iterate: bad to better

IterateLadder walked you up short rounds: notice what’s wrong, tighten the ask, climb again. You practiced editing the prompt instead of blaming the model. That habit beats hunting for a mythical “perfect starter prompt” forever.

Round 1 often fails for a clear reason: too vague, wrong format, missing constraint, too long, or the wrong tone. Name the failure in one line, then adjust only that.

Example ladder — reply email:

Round 1 (weak):

Write an email about missing the meeting.

Round 2 (add job + context):

Write a short apology email to my teammate Sam.
I missed the 10am product sync. I reviewed the notes.
Ask if anything still needs my input before Thursday.
Tone: accountable, not dramatic. Under 100 words.

Round 3 (fix one complaint):

Same email, but cut the apology to one sentence
and lead with what I’m doing next. Keep under 80 words.

Same ladder works for dinner (“less dairy” → “skillet only” → “under $12”) and study (“more quiz-like” → “hide answers” → “harder on chapter 4”). Don’t restart from zero each time — answer the model with: “Keep everything else; change X.”

• First draft is “fine” but not sendable
• Format is right, facts or tone are off
• You’re close and don’t want to rewrite the whole prompt

Thrashing — changing five things at once — hides what helped. Also don’t iterate forever on facts that need a real source; after two rounds of polish, verify anything important yourself.

Save a bad prompt and its reply. Improve it twice with one change each. Keep both versions so you can see the climb.