Skip to content
anythingai.space
Prompting & Contextbeginner11 min · 9 steps

Prompt Engineering Fundamentals

Treat a prompt like code: state the intent, fence the data, respect the instruction hierarchy, and get consistency from the system rather than from temperature.

run 1run 2run 3Same prompt, three runs, three differentshapes.Prompt“Help the user.”LLMAnswer A3 chatty paragraphsAnswer Ba bullet listAnswer Casks a question back
requestresponseerrortool calldata
1 / 9
ProblemStep 1 of 9

A vague prompt gets a different answer every time

“Help the user with their request.” The model has to guess what help means, how long to be, what format to use and what to avoid. Each run it guesses differently.

Nothing is wrong with the model. The prompt never said what a finished answer looks like, so the output has no reason to stay the same.

Watch outIf the prompt doesn't define success, every output counts as success, and none of them can be tested.

Lesson 1 of 8 in Prompting & Context