Indictment: A template wearing a prompt's clothes — every bracket is a blank check for failure, and '95% certain' is a made-up metric that tells the model nothing.
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.
“top 0.5% expert in [field of expertise]”
The bracket is a placeholder the user forgot to fill in, so the model has zero domain context and will hallucinate a generic expert persona.
swap: '[field of expertise]' → the actual field, e.g. 'B2B SaaS pricing strategy'
Never ship a prompt with unfilled brackets — placeholders are not prompts, they're homework you assigned yourself.
“I want to [goal]. Provide options with trade offs for each.”
'[goal]' is literally empty, and 'options with trade-offs' with no scope means the model decides how many options, how deep, and what format — all wrong.
add: specific goal + 'Return exactly 3 options, each with a one-line summary, 2 pros, 2 cons, and a recommended use case'
Whenever you ask for options, define the count, structure, and depth — otherwise the model fills every gap with its own defaults.
“Ask me any questions until you are 95% certain”
'95% certain' is an unmeasurable pseudo-metric that the model will interpret however it likes, producing either one clarifying question or twenty.
swap: '95% certain' → 'Ask me up to 3 clarifying questions, one at a time, before proceeding'
Replace fuzzy confidence thresholds with concrete behavioral rules: a question count, a stopping condition, or a specific list of what to clarify.
“Here are the [background and context].”
Another unfilled bracket — the model receives no actual context, so it will either stall on clarifying questions or invent assumptions.
cut: the bracket entirely and replace with the real background inline, or instruct the user to paste it before sending
Context fields must contain actual content — if you're building a reusable template, write explicit fill-in instructions, not silent brackets.
You are a senior [FILL: specific role, e.g. 'SaaS pricing strategist with 15 years of B2B experience']. My goal: [FILL: one concrete sentence, e.g. 'Choose a pricing model for a new API product targeting mid-market engineering teams'] Background: [FILL: 3–5 bullet points — company stage, constraints, what you've already tried, what success looks like] Before generating anything, ask me up to 3 clarifying questions, one at a time. Stop asking once you have enough to proceed. Then deliver your response in this format: - **Option [N]: [Name]** - Summary: one sentence - Pros: exactly 2 bullets - Cons: exactly 2 bullets - Best for: one sentence describing the scenario where this wins Provide exactly 3 options. No preamble, no conclusion paragraph. Total response under 500 words.
https://getmeerkat.dev/roast/r9x5j2fm8k