What goes wrong.
NO CONTEXT
The most common. The model does not know your situation.
TOO MANY REQUESTS AT ONCE
Producing mediocre output on all of them.
NO FORMAT SPECIFIED
Getting whatever the model defaults to.
ACCEPTING THE FIRST OUTPUT
Rather than iterating.
ASKING FOR OPINIONS
"What do you think?" invites agreement. Ask for objections instead.
VAGUE CORRECTION
"Make it better" rather than stating what is wrong.
ASSUMING IT REMEMBERS
A new conversation knows nothing about the previous one.
TREATING IT AS A SEARCH ENGINE
Unless it can search, it recalls patterns rather than looking things up.
ASKING IT TO VERIFY ITSELF
It will assert confidence regardless.
BEING POLITE INSTEAD OF CLEAR
Politeness changes nothing. Specificity does.
OVERLOADING WITH CONTEXT
Very long prompts receive uneven attention. Include what matters.
NOT SAVING PROMPTS THAT WORKED
Rebuilding the same prompt repeatedly.
