Diagnosis: Solid bones, but 'options with trade-offs' is shapeless and '95% certain' is a number the model can't measure.
see the specimen they pasted
You are a top 0.5% expert in [field of expertise]. I want to [goal]. Provide options with trade offs for each. Here are the [background and context]. Ask me any questions until you are 95% certain you understand my asks. Do not generate the response until you do.
“Provide options with trade offs for each”
No count, no shape — the model will return 2 options or 8, with one-line trade-offs or five paragraphs each, and all of it technically satisfies this instruction.
swap: 'options with trade offs' → 'exactly 3 options; for each: 2 pros, 2 cons, and one sentence on who it's best for'
Pin every list output: count of items, internal structure per item, and the deciding criterion — otherwise the model picks all three at random.
“Ask me any questions until you are 95% certain”
A model has no internal certainty meter — '95% certain' is unmeasurable, so it either asks one question and proceeds or spirals into an interrogation with no stop rule.
swap: 'until you are 95% certain' → 'ask up to 3 clarifying questions, one at a time, then proceed'
Replace self-assessed model states ('until certain', 'when confident') with concrete stop rules: a question cap, a round limit, or a specific trigger condition.
“top 0.5% expert in [field of expertise]”
Percentile framing is decoration — it tells the model nothing about how that expert thinks, communicates, or what they prioritize; the role is hollow without a behavior signal.
add: a behavior cue after the role, e.g. '[field of expertise] expert who cuts to trade-offs fast and flags hidden assumptions'
A role without a behavior cue is a costume — add one sentence on how the expert thinks or communicates to make the persona functional.
You are a world-class [field of expertise — e.g. 'cloud infrastructure architect' or 'corporate tax strategist']. You cut to trade-offs fast, flag hidden assumptions, and don't pad answers with caveats. My goal: [goal — one concrete sentence, e.g. 'choose a database architecture for a multi-tenant SaaS product'] Background and context: [background and context — include constraints, existing decisions, and any non-negotiables] Before responding, ask me up to 3 clarifying questions — one at a time — to fill any gaps. Once you've asked them all (or confirmed you have enough), produce the following: **Exactly 3 options.** For each: - Option name + one-line summary - 2 pros (specific to my context) - 2 cons (including hidden or non-obvious ones) - Best for: one sentence on when this option wins Keep the full response under 600 words. No filler, no preamble.
https://getmeerkat.dev/roast/eds9kcytg8