You Don’t Know What You’re Asking For
Most bad AI answers start with a question the person hadn’t actually decided yet.
Before a prompt needs better formatting or context, it needs something earlier: a decided question. Most people skip this without noticing, because typing the question and figuring it out have quietly become the same step.
“Can you help with this function, it’s not working right” isn’t vague because of weak prompt-writing — it’s vague because “not working” hasn’t been decided yet. Error? Wrong output? Slow? Each is a different task. The model can’t untangle what the person hasn’t, so it guesses, and the real question gets sorted out a few exchanges later, after a couple of misses that felt like normal back-and-forth.
Compare: “this function should return null for negative input, and it’s returning zero — find why.” Not faster to write. Slower, actually — the time went into deciding, not phrasing.
The test before you type: what, specifically, am I asking — not the general area, the actual question. If that doesn’t fit in one sentence, you’re not ready to ask, and any answer you get will be guessing at the same thing you are.
For an individual, this comes before format or context — they only matter once the question is decided. For a business, this is usually the real story behind “the AI gave a useless answer”: the question was never decided before the conversation started doing both jobs at once.
— Arvind, Rationale One short issue a week. No jargon, no hype — just the reasoning behind what’s changing.