Steal the 5-line brief.fill in the blanks, hand over the whole job.
Anthropic shipped its smartest model on June 9. Its lead grows with task length — so the skill isn’t clever prompting anymore. It’s a clear brief. This is the full manual: the template, the prompts, and the two other moves most people will scroll past.
Why briefs beat prompts
The receipt: Stripe pointed Fable 5 at a 50-million-line codebase. A migration scoped at two-plus months for a whole engineering team shipped in one day (Anthropic launch announcement, June 9, 2026). That’s not a chatbot — that’s staff. And staff doesn’t run on one-line questions. It runs on briefs.
Here’s the mechanism, in Anthropic’s own framing: the longer and more complex the task, the larger Fable 5’s lead over earlier models. Short prompts can’t carry long tasks — they starve the model of context, constraints, and a definition of done. A brief carries all three. The model also works autonomously longer than any previous Claude, which means the unit of work you hand it should be a job, not a question.
The 5-line brief — fill in the blanks
Why each line earns its place
- Job kills the topic-prompt. “Write about SEO” gets you filler; “audit why this page stopped ranking” gets you work.
- Context is where most people leave 80% of the output quality on the table. Attach the actual files — the model reads them.
- Constraints are cheaper than revisions. Say what’s off-limits before, not after.
- Done = is your acceptance test. If you can’t write it, you’re not ready to delegate it.
- Effort maps to a real Fable 5 behavior: at the highest effort setting it reflects on and validates its own work before showing you. Ask for it on anything that matters.
Copy-paste prompts for real marketing jobs
Don’t admire the framework — run it. Three filled-in briefs you can paste today. Swap the bracketed bits for your business.
JOB: Audit why our [/services/seo/] page stopped ranking for "[target keyword]" and write the fix list. CONTEXT: [Search Console screenshot — 3-month performance] [GA4 landing-page screenshot] [the page URL] [top 3 competitor URLs] CONSTRAINTS: Our voice doc is attached. No new tools, no rebuild — work with the page we have. No invented stats. DONE = A prioritized fix list I can hand to my designer, with the why behind each item and the expected impact, ranked. EFFORT: Check your own work against the screenshots before you show me.
JOB: Build my next 30 days of social content, ready to schedule. CONTEXT: [voice/brand doc] [last month's 10 best posts with engagement numbers] [my offer + who it's for] [2 competitor accounts that are landing] CONSTRAINTS: My voice, no hustle language, no hashtags. Each post = one idea. Mix formats: carousels, single images, text posts. DONE = A calendar table: date, hook, format, caption draft, CTA. Ready to schedule without me rewriting it. EFFORT: Highest — review the calendar against my best performers before you show me.
JOB: Turn my last 90 days of Google reviews into marketing assets. CONTEXT: [screenshot or export of Google Business reviews] [my services list] [one example of my writing voice] CONSTRAINTS: Quote customers verbatim only with light cleanup. Flag anything that needs a reply from me first. DONE = (1) the exact customer phrases I should use as headlines, (2) 5 social posts built on real reviews, (3) reply drafts for any review under 4 stars. EFFORT: Quick pass is fine — I'll pick winners.
Give it a notebook — memory files
The sleeper feature in the launch notes: persistent, file-based memory improved Fable 5’s long-task performance three times more than the previous model in Anthropic’s tests. Translation for operators: the model gets better at YOUR business the more structured material it can keep and consult.
The minimum viable setup is three files. Start the brand file from this skeleton:
BUSINESS: [name, what you sell, where, since when] CUSTOMER: [who actually buys — be specific, name the niches] OFFER: [your core offer + price posture] VOICE: [3 adjectives + 3 banned words/phrases] PROOF: [your 3 best receipts — results, reviews, numbers] GOALS THIS QUARTER: [1-3, measurable] NEVER: [claims, topics, or tactics that are off-limits]
Add a voice file (5 real sentences you’ve written that sound like you, 5 that don’t) and a past-work folder (your best outputs). Reference them in every brief’s CONTEXT line — and let the model keep its own notes between sessions. That’s the 3× compounding.
Stop describing. Start showing.
Fable 5’s vision is state-of-the-art — in Anthropic’s demos it rebuilt a working app from screenshots alone. For marketers, that flips the workflow: your dashboards are prompts now. Don’t summarize what Search Console says — screenshot it.
JOB: Tell me what these dashboards say that I'm not seeing. CONTEXT: [Search Console performance screenshot] [GA4 acquisition screenshot] [screenshot of the SERP for my main keyword] CONSTRAINTS: Specifics only — name the queries, pages, and trends in the images. No generic advice. DONE = The 3 most important things in these screenshots, each with: what it is, why it matters, what I should do this week. EFFORT: Check your reading of the numbers twice before you show me.
Ranking isn’t the prize anymore. Getting cited is.
Models this good are where buyers research now. Answer engines collapse the old journey — the user asks, the model answers, no click. A page can sit at position one in Google and still never get quoted by Claude, ChatGPT, or Perplexity if a machine can’t lift the answer from it.
The citation-ready checklist we run on client pages:
- Answer in the first line of the section, not after three paragraphs of wind-up.
- First-hand proof — a number, a result, a named example AI can’t invent on its own.
- Clear claims with dates and sources — models prefer quotable, attributable statements.
- Schema + anchored sections so the structure is machine-parseable (this page practices what it preaches).
This is the lane we work in every day — see our Hawaii SEO services and the full Hawaii SEO guide for how citation-readiness fits a complete strategy.
The fine print everyone skipped
Fable 5 is included on paid Claude plans only through June 22, 2026. After that it moves to usage credits ($10/M input, $50/M output tokens via the API — under half of what the Mythos preview cost, but no longer bundled free). Two weeks to learn the manual while the meter’s off. Run one brief on your messiest marketing job and see what a day’s work looks like now.
FAQ
What is Claude Fable 5?
Anthropic’s most capable generally available AI model, released June 9, 2026. It’s the first of the “Mythos-class” tier — previously restricted capability now open to anyone on a Claude plan — and its advantage over earlier models grows as tasks get longer and more complex.
What’s the 5-line brief?
A fill-in-the-blank delegation template: Job, Context, Constraints, Done =, Effort. It replaces prompt tricks with the same structure you’d use to brief a contractor — which is how models at this level do their best work.
Do I need the API or coding skills to use this?
No. Everything on this page works in the regular Claude app. The brief is pasted as a message; context files are attached; screenshots are dropped in.
What happens after June 22, 2026?
Fable 5 stops being bundled free with paid Claude plans and runs on usage credits instead. The workflows on this page still work — you’ll just be paying per use, so the free window is the cheapest time to build the habit.
Want us to run it for you?
We’re a Wailuku, Maui team that ranks, builds, and runs paid traffic for businesses on all four islands — and we’ve already put Fable 5 to work. One free consult, zero pitch.
