Prompt of the day

Appointment brief: the questions to ask before the door closes

You get one short appointment. Walk in with a structured brief instead of a vague worry — and walk out knowing what happens next.

You are helping me PREPARE for a medical appointment. You are not diagnosing me and you must not try. I will describe what's going on and what I want from the visit. Produce:

A) THE 60-SECOND SUMMARY — my situation as I'd say it out loud to a clinician in one minute: what, when it started, how it's changed, what makes it better or worse, what I've already tried. Use only what I told you.

B) THE QUESTION LIST — the questions worth asking, ordered by what would most change what I do next. Include the three that patients most often forget: what else could this be, what happens if we do nothing, and what would make you change your mind.

C) DECISION POINTS — for any test or treatment likely to come up: what the result would actually change. If a result wouldn't change anything, that's worth knowing before I agree to it.

D) WHAT TO WRITE DOWN — the specific things to capture during the visit (names, doses, next steps, when to follow up, what warrants coming back sooner).

E) URGENT-CARE CHECK — based on what I described, anything that should NOT wait for this appointment. Be conservative: if in doubt, say go now.

Inputs: [WHAT'S GOING ON + TIMELINE] · [WHAT I'VE TRIED] · [WHAT I WANT FROM THIS VISIT] · [RELEVANT HISTORY I'M COMFORTABLE SHARING]

Rules: Do not diagnose, do not name a likely condition, do not suggest medications or doses. You are building a communication tool, not practicing medicine. Do not ask me to paste medical records, test results with identifiers, or insurance details. If what I describe could be an emergency, say so first, before anything else. Do not invent facts, sources, citations, links, or specifics you cannot support — say "I'm not sure" instead, and mark anything I must verify for myself rather than asserting it.

Why it works — The failure mode in a 12-minute appointment is not the doctor — it's an unstructured patient. This prompt does the one thing an LLM is genuinely good at here (structuring what YOU know into a brief and a question list) and is hard-blocked from the thing it's dangerous at (diagnosing).

Catch AI's mistakes by asking what it assumed

Confident-but-wrong is the #1 AI risk at work. Surfacing its hidden assumptions is a 10-second insurance policy on every important output.

After any answer you'll act on, ask: 'What did you assume that I didn't tell you? What would change your answer if it were wrong?'

1. Get the answer.
2. Ask for its assumptions + what would flip them.
3. Verify the load-bearing ones.