The quality of an AI model's output depends heavily on the quality of your input. This is a real, learnable skill - not guesswork.
Be specific about the output format
# Vague:
"Tell me about Python"
# Specific:
"Explain Python in 3 bullet points for someone who has never coded before,
avoiding jargon."
The second prompt tells the model the length, the format, and the audience - all things it would otherwise have to guess.
Give the model context it does not already have
# Missing context:
"Fix this bug"
# With context:
"This Python function should return the average of a list, but returns 0
for a list of [1, 2, 3]. Here is the code: [paste code]. What's wrong?"
Break complex tasks into steps
Asking for one giant, complex thing in one shot often produces worse results than guiding the model through smaller steps - the same way breaking a program into functions makes it more reliable:
# One giant ask:
"Build me a complete e-commerce website"
# Broken into steps:
"First, list the main pages a simple e-commerce site needs.
Then we'll design the database schema.
Then we'll build one page at a time."
Common mistakes
Assuming the model remembers something from a previous, separate conversation (it does not, unless that context is explicitly included). Accepting the first answer without verifying anything important - covered further in the Limitations lesson. And over-trusting confident-sounding output on topics where accuracy actually matters, like medical, legal, or financial specifics.