I’ve spent the past few days testing Claude Fable 5, Anthropic’s new top-of-the-range model, and I’ll give it to you straight.
It felt like a casino.
Some sessions, ding ding ding, jackpot! It planned like a strategist, asked me questions I hadn’t thought to ask myself, and handed back work I would have happily paid a consultant for.
Other sessions? Like ‘letting it ride’ on a roulette table. Watching credits disappear on an answer a cheaper model would have given me for free.
So I went looking for the pattern because “sometimes it’s amazing” isn’t a system, and you know how I feel about systems.
Before you write Fable 5 off as expensive roulette (or burn your credits discovering this yourself), let me show you what changed my results – and give you some real estate prompts to try yourself.
First, what you’re dealing with
Fable 5 sits at the top of Anthropic’s range, above Haiku, Sonnet, and Opus.
What makes it different from every model before it is that it plans before it writes, asks you questions before it starts, works through long, multi-step jobs without losing the thread, and checks its own work before handing it back.
Like an employee who re-reads the email before hitting send.
The catch is the price tag: For the next three days (until July 12), it’s included in paid Claude plans, but then moves to pay-per-use: US$10 per million input tokens and US$50 per million output tokens.
And because the model re-reads your entire conversation at every turn, long chats compound fast.
Years ago, when I started speaking about AI, I made the comment that AI is the new electricity, like something we will eventually switch on and pay per use. Fable is like that.
But smarter AI is expensive AI. So my message today: stop hunting for the best prompts. Start deciding which of your problems deserve the expensive brain.
Don’t book Heston to make a toasted cheese
Claude Fable is a bit like the partner of an expensive accounting firm. The partner does the consulting and the juniors do much of the work (it keeps the costs down, they tell me). The partner charges $ 1,000 per hour, and the junior charges $100 per hour.
You know I’m a MasterChef fan, so let me put it this way: you wouldn’t book Heston Blumenthal to make you a toasted cheese sandwich.
Could he? Of course. It would be the best toastie you’ve ever had. There would probably be a smoked applewood foam involved. But you’d be paying degustation prices for something the sandwich press does perfectly well.
You book Heston to design the menu. To rethink the whole kitchen. To taste what you’re serving and tell you what your restaurant should become.
Then someone else does the toasting of the bread.
Fable 5 is Heston. Sonnet and Opus are your line cooks, still included in your plan, and better than anything any of us were using six months ago.
In real estate terms: “rewrite this OFI follow-up SMS” is a toastie. Don’t send it to Fable.
The jobs to send it are the ones you’d happily pay a coach or consultant a few hundred dollars to think through with you.
Designing your 90-day client and prospect nurture system. Stress-testing whether to go auction-first in your market. Turning the raw exhaust of your week into an operating plan.
How to brief a $1,000-an-hour brain
Anthropic published guidance on prompting Fable 5, written for developers, so here it is basically translated.
Strip it back, and four lines changed my results more than anything else.
Lead with the outcome. Your first sentence should say what you need, before any backstory. “I’m pitching a $3 million listing on Thursday, and I need to walk out with a soft yes, alignment on method of sale, and an agreed marketing budget” beats “Can you look at my listing presentation and give me some ideas?” every time.
Demand evidence. Tell it: only make market or pricing claims you can tie back to the data I’ve supplied, and if you’re making an inference, label it as an inference. This one line keeps Fable out of the invented-statistics business and keeps your vendor reports defensible.
Say whether you want advice or action. Fable is eager. Ask for feedback on an email, and it will cheerfully write the email. So declare your mode: “I’m thinking out loud about auction versus private treaty for this property – thoughts only, don’t draft anything yet.” Or: “Handle this end-to-end and only check in if something needs my judgement.”
Make it commit. “I’ve told you my priorities. Give me one recommendation, then the strongest argument for the option you rejected.” No more “it depends”. You’ve told it what you value; hold it to a verdict.
Mostly, that’s the difference between the ‘jackpot’ sessions and the expensive shrugs.
Start here: the weekly operating review
Of everything I tested, this is the prompt to run first, because it solves the thing I hear from agents and principals every week: “I don’t have time to learn all this stuff because I am too busy.”
Hand Fable the raw notes of your week – listing appointment notes, OFI feedback, WhatsApp threads, email snippets, the to-do list on the back of an envelope – and ask it for an operating plan.
Copy this one exactly:
I want you to turn the raw notes from my week as a real estate agent into a simple operating plan I can run next week from. The goal: nothing falls through the cracks – especially commitments to vendors and buyers.
I'll paste in: listing appointment notes, OFI feedback, WhatsApp and SMS threads, email snippets, and my to-do lists.
My top priority for the next two weeks is: [describe].
Only include actions or decisions you can tie back to something in my notes or messages. If you're inferring, label it as an inference. Don't invent extra "good ideas" to fill the page.
You're advising, not acting. Don't draft any emails or SMS yet – tell me what needs to happen and why. I'll ask for scripts separately.
Done means one document with four sections:
- Commitments – everything I promised vendors, buyers, my principal, or my team, with who / what / when. Flag anything overdue.
- Decisions – what was decided this week (pricing, marketing, strategy) versus what was only discussed. Ambiguous items go in an "unclear" bucket.
- Conflicts – anywhere two conversations point in opposite directions. Quote both.
- Next week – the five actions that most advance my stated priority, each traced to a note, email, or conversation.
When you have enough information, commit. One recommended plan, plus the strongest argument for a different focus if you think I should consider one.
That conflicts section is the sleeper. It catches the vendor who was “in no rush” on Tuesday and “must be sold before school goes back” on Friday – in their own words, quoted side by side.
A few other Fable-Worthy jobs to try
Stress-test a strategy decision
I'm about to make a significant decision in my business and I want it stress-tested before I commit, because reversing it later will be expensive in time, money, and credibility. Treat this like the review a sharp board member would run.
The decision: [e.g. moving my core market to auction-first versus staying with mixed methods for the next 12 months].
First, ask me for: the options on the table, my priorities ranked in order, the constraints I can't change, and the deadline. Then interview me – a few questions at a time – until you understand the decision better than I've articulated it.
Push on anything I'm vague about; vagueness is usually where the risk hides.
Once you have enough, lead with your recommendation in one sentence. Then show:
- How each option scores against my ranked priorities
- Best case, base case, and worst case for your recommended option
- The strongest argument for the option you rejected
- The single most likely way your recommendation turns out to be wrong
Be direct. Don't hedge with "it depends on what you value" – I will have told you what I value. If my priorities contradict each other, name the contradiction.
Expand your market share
I need a research brief I can make a decision from about expanding into
[suburb/region] – a decision-grade document, not a generic suburb profile.
Wrong claims are worse than missing claims.
Start by asking me: the decision this informs (second office, marketing into the area, or leaving it for now), the timeframe, and the data sources I already use and trust.
Then research end to end:
- Stock volume and turnover over the last 12–24 months
- Price points and key segments
- Who dominates listing share and what they appear to be known for
- Visible gaps or underserved segments
Open with the answer in three sentences. Cite a source for every substantive
claim. Where sources disagree, say so and tell me which you'd trust and why.
Separate evidence from inference, and label the inferences.
Before delivering, attack your own draft: challenge the three claims your
recommendation most relies on, as if you were a sceptic paid to find the flaw.
Keep it under two pages – cut anything that wouldn't change what I do next.
Turn your raw notes into a market update
I want to develop a piece of writing from raw material – OFI notes, vendor
conversations, buyer questions, and my own dot points – where the argument hasn't taken shape yet.
Your job is to find the story inside this material, not to write a generic "state of the market" piece.
Ask me to paste the raw material, then ask: who this is for (buyers, sellers,
investors), where it will be published, and what I want readers to think or feel by the end.
Then, before drafting:
- Identify the most interesting tensions, surprises, and unresolved questions
in the material
- Ask me about the judgement calls only I can make – what I believe, what I've seen on the ground, what I'm willing to say publicly
- Propose three different angles this piece could take, with one line each on
what that version emphasises and what it sacrifices. Then stop and wait for
me to choose.
After I choose, draft in my voice – my phrasing, my rhythm. Flag every assertion my material doesn't support so I can supply evidence or cut it. Don't invent quotes, anecdotes, or numbers; leave a visible gap instead.
Marketing plan – Fable sets the strategy, your cheaper models write the copy:
I'm preparing to launch a new listing and want you to act as my listing launch co-pilot – strategy first, copy later.
Property: [suburb, beds/baths, key features]. Target buyers: [describe].
I'll paste in photos, floorplan details, and past campaign material.
Constraints: I'm an Australian agent, so everything must sit within advertising and fair trading rules – no invented features, no underquoting, nothing misleading. Tone must match my brand: [describe].
Your job:
- Design the overall launch strategy and channel mix
- List exactly which assets I need (portal copy, email, social, video, print)
and what each one has to achieve
- Write a short, clear brief for each asset that I can hand to a cheaper model(or a human) to execute
Lead with your recommended strategy in one paragraph. Ask your clarifying
questions before you lock the plan. Don't write the full copy yet.
Keeping the bill under control
Try those prompts while you can, but if you’re reading this post July 12, here are some of the things I’ve learned this past week:
New chat for every project. One listing launch, one operating review, one decision per conversation. Run your whole month in a single thread, and you pay for Fable to re-read your history at every turn (message).
Two or three Fable turns, then switch. Use Fable to frame the problem, make the call, and outline the plan. Then drop to Sonnet or Opus – still included in your plan – to fill the templates and grind out the variations.
Principals: write the policy down. Fable is for strategy, system design, operating reviews, and decisions. Sonnet and Opus are for listing copy, emails, social posts, and scripts. If a good assistant could do it in twenty minutes, the expensive brain doesn’t get it. If you don’t have an AI Policy yet, get a draft here.
And one line on compliance, because it matters: Fable can help you think about what you’re allowed to say, but checking it against your state’s rules is still your job. The buck stops with the human holding the licence, who is supposed to protect clients’ privacy.
Your move this week
Run one experiment before the weekend. After that, the API meter starts.
Gather the stuff from your week – the notes, the threads, the feedback, the envelope – paste it into the operating review prompt above, and see what comes back.
If it feels like a pull of a lever without the “kaching”, you did not brief it well enough.
If it feels like the sharpest operations manager you’ve never hired just organised your fortnight… now you know which table you’re sitting at, and what the stakes are worth.
And lastly, for the love of everything: Don’t ask Fable to rewrite your email to make it “sound better”. It’s the digital equivalent of booking Heston Blumenthal to make you a toasted cheese.
Note: Members of Elite Agent Insiders got these prompts yesterday. Join insiders here.