Diagnosis: Solid skeleton, but 'options with trade-offs' and '95% certainty' are structural holes that survive every fill-in.
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 returns 2 options or 9, a bullet or a table, and the user has no grounds to complain because they never specified.
swap: 'options with trade offs' → 'exactly 3 options; for each: 2 pros, 2 cons, and one sentence on who it suits best'
Always pin list outputs to a count and a per-item shape — 'options with trade-offs' is an instruction, not a format contract.
“Ask me any questions until you are 95% certain”
A model has no instrument to measure its own certainty — it either fires one token of clarification or spirals into an interrogation with no exit condition.
swap: 'until you are 95% certain' → 'ask up to 3 clarifying questions, one at a time, then proceed'
Replace unmeasurable confidence thresholds with concrete stop rules: a question cap and a trigger to proceed.
“top 0.5% expert in [field of expertise]”
Percentile framing is decorative — the model has no calibration for '0.5%' and defaults to generic expert voice anyway; the role needs a behavioral signal, not a ranking.
swap: 'top 0.5% expert' → '[field of expertise] expert who defaults to first-principles reasoning and flags assumptions explicitly'
Give the model a behavior to perform, not a rank to inhabit — roles work when they specify how to think, not how good to be.
You are a [field of expertise — e.g. 'corporate tax attorney', 'SaaS growth strategist'] with deep first-principles expertise. You flag assumptions explicitly and default to structured, decision-ready output. My goal: [goal — one concrete sentence, e.g. 'choose a pricing model for a new API product'] Background and context: [background and context — key constraints, stakeholders, timeline, prior decisions] Before responding, ask up to 3 clarifying questions — one at a time — to resolve anything that would materially change your answer. Once you have what you need (or after 3 rounds), proceed. Deliver exactly 3 options. For each option: - Name and one-line description - 2 pros (specific, not generic) - 2 cons (specific, not generic) - Best suited for: one sentence on when or for whom this option wins Close with a single recommendation and the one sentence rationale behind it. Total response: under 500 words.
https://getmeerkat.dev/roast/3x4cuyykj9