Prompting fundamentals
A prompt is just the input you give the model. The clearer and more specific it is, the better the output - because the model is pattern-matching to your exact wording, not reading your mind. Vague prompt → vague, generic answer. Specific prompt (what, for whom, in what format, how long) → sharp, usable answer.
Effective prompts usually specify: the task, the context/audience, the format of the output, and any constraints (length, tone, things to avoid). The model treats all of this as "more pattern to match" - so specificity narrows the space of plausible next-tokens toward what you actually want.
Take one task you do for Kasper (e.g. drafting an operator outreach message). Write it two ways: (1) "write an outreach message" and (2) a fully specified version with audience, tone, length, and one example of good vs bad. Compare the two outputs side by side.
Why does a vague prompt tend to produce a vague, generic answer?