Vague prompts produce generic answers because the model optimizes for the average of all possible contexts. Specificity forces it to optimize for your situation.
The fastest improvement: replace every adjective with a measurement. Replace every assumption with an explicit statement.
02 Weak vs. Strong
EX 01Medical concept for a specific patient
Explain Type 2 diabetes to a 52-year-old who was just diagnosed. They have a nursing background and understand basic physiology (blood glucose, insulin, pancreas function) but have never studied endocrinology.
Length: 3 focused paragraphs.
Tone: honest and direct — they can handle clinical language, they just want clarity, not cheerful minimizing.
Must cover: mechanism, why it progresses over time, and the 3 most important lifestyle levers.
Do NOT use the phrase 'lifestyle disease' — they find it stigmatizing.
End with: one thing most newly-diagnosed patients wish they had known earlier.
→ Why it works
Audience defined precisely (nurse, just diagnosed, specific knowledge level). Tone described including what to avoid.
EX 02Landing page headline for a real product
Write 5 landing page headlines for Vitae — a resume tool that shows you analytics after you apply: who opened your resume, how long they spent on it, which sections they focused on.
Target buyer: product managers and engineers with 5–12 years of experience, quietly job searching while employed.
Tone: confident and data-driven. NOT motivational, NOT hustle-culture.
Format: each headline on its own line, max 8 words, no punctuation at end.
Avoid: take control, stand out, land your dream job, get noticed.
Requirement: one headline leads with the analytics angle, one leads with the insight angle, one is a direct question.
→ Why it works
Product and differentiator specific. Target persona precise (quiet job searchers, 5–12 years).
03 Key Points
01Not 'short' — '3 sentences' or 'under 80 words'
02Not 'for beginners' — 'for a paralegal who has never used Python'
03Not 'professional' — 'like a McKinsey consultant emailing a CFO'
05One clear goal per prompt — chain multi-goal tasks
04 Model-Specific Notes
Claude infers intent well but explicit always beats implicit. 'Professional' vs 'direct' vs 'warm but authoritative' produce meaningfully different outputs.
05 For Your Role
After writing your prompt, ask: how many different valid answers could this have? If more than 3, add constraints until it's 1–2.