Loading...
Prompting the new Claude is drastically different than the old one. 7 rules from Anthropic's official 31-page guide, distilled into what actually matters.
Old Claude (4.6)
New Claude (Opus 4.7)
Opus 4.7 does exactly what you type. No more guessing what you meant.
Old prompt
Review this contract.
New prompt
Review this contract. Flag risks per clause. Rate severity 1-5. Suggest one rewrite per risky clause. Return as a table.
The fix: Name every output. Name the order. Name the boundaries.
Opus 4.7 sizes the answer to what it thinks the task is. Long input + "summarize" = long summary.
Old prompt
Summarize this report.
New prompt
Summarize this report in exactly 5 bullet points. Each bullet under 15 words. First word of each bullet: an action verb.
The fix: Name the format and the cap.
Negative instructions stick to the literal sentence. They don't work on 4.7.
Old prompt
Don't use jargon. Don't use buzzwords. Don't sound like a marketer.
New prompt
Write in plain English a 16-year-old could read aloud. Use short, concrete words: simple, specific, real. Replace "leverage" with "use." Replace "scalable" with "works at any size."
The fix: Tell Claude what TO do, not what NOT to do.
Each action verb tells Opus 4.7 to ship something specific. Vague requests get vague results.
Old prompt
Can you help me with the email?
New prompt
Write the answer email. Final draft. Send-ready. Goal: book a meeting with the CRO by Friday. Length: under 90 words. Tone: confident, casual, specific.
The fix: Start with a verb. Add constraints. Define done.
Opus 4.7 calls fewer tools by default. It reasons more between calls.
Old prompt
(No explicit tool instruction)
New prompt
Use web search aggressively. Verify every claim with at least 2 sources.
The fix: If quality is good, trust the default. If you need more tools, say so explicitly.
Opus 4.7 is more direct. Less validation. Almost zero emojis. No more "Great question!" energy.
Old prompt
(Default 4.7 tone: direct, minimal)
New prompt
Use a warm, conversational tone. Acknowledge the user's framing before answering. Match the rhythm of these examples: [paste 2-3 sentences in your voice].
The fix: Paste examples of your voice. Tell Claude to match the rhythm.
Opus 4.7 hits the literal minimum on open-ended work unless pushed.
Old prompt
Build a landing page for my AI consultancy.
New prompt
Build a landing page for my AI consultancy. Sections (in order): Hero, Logo bar, 3 case-study cards, Service blocks, Testimonial carousel, About me, Newsletter signup, Footer. Style: editorial, serif headlines, sans-serif body, generous whitespace. Animations: subtle on scroll. No purple gradients. Go beyond the basics. Polish like it's a real client deliverable.
The fix: This phrase is from Anthropic's own Opus 4.7 documentation. It pushes past the literal minimum.
Pro trick
Opus 4.7 uses "adaptive thinking" — it doesn't reason by default. Add this phrase at the end of any prompt to force maximum reasoning:
Think before answering (maximum reasoning)This forces Claude to use extended thinking on every response. Much better answers.
| Do | Don't |
|---|---|
| Name every output, order, boundary | Use vague verbs like "review" or "help" |
| Define format and length cap | Say "summarize" without constraints |
| Use positive instructions ("write in plain English") | Use negative instructions ("don't use jargon") |
| Start with action verbs | Start with "Can you help me" |
| Paste voice examples for tone | Expect warmth without asking |
| Add "go beyond the basics" for creative work | Expect creative polish by default |
| End with "Think before answering (maximum reasoning)" | Assume adaptive thinking is always on |
AI Brain Pro has all 7 prompting rules baked into the operating system. Claude follows them automatically on every task. Plus 138 skills, 110 agents, and the HALO self-improvement loop.
Get AI Brain Pro — $97