How to Make Money With Fable 5: Greg Isenberg's Full Playbook
Last week I wrote about the weekend Claude Fable 5 went dark and everyone panicked.
This is the sequel. Fable came back, and Greg Isenberg sat down for 33 minutes to answer the only question that matters: how do you actually make money with Fable 5 instead of just watching people one-shot Monopoly games on X?
His opening line sets the tone: "99% of people are using Fable 5 wrong."
They're benchmarking it. They're screenshotting it. They're not pointing it at problems that print money.
Greg's video (86K views and counting) is the most tactical breakdown I've seen of what this model unlocks. Prompts included. Startup ideas included.
Here's everything worth stealing from it.
The Anthropic launch video that cost $0 in editors
The example that made Greg record this episode: a guy at Anthropic edited and shipped his own launch video using Fable.
Not a rough cut. A finished, color-graded, animated launch video. The kind of thing Greg says would normally cost hundreds of thousands of dollars.
The setup was simple:
→ 17 takes across 4 scenes, dumped in a folder
→ One prompt telling Fable to transcribe everything with ElevenLabs, pick the best takes (usually the last one, fewest ums), and stitch them into a final clip
→ FFmpeg executes the edit from a JSON file. Remotion rebuilds every design frame as code
→ The prompt ends with: "Don't stop until you have the final video"
And it worked. Even the color grade. When it came back muted, he typed "the color grading feels a bit muted, can you fix that" and it fixed it.
And the lesson goes past "AI can edit video now."
Most of us are vibe coding apps nobody sees. The same model can produce the distribution assets that get people TO the app.
"Low effort is the alpha"
My favorite counterintuitive bit in the whole video comes from Greg's friend Morgan Linton:
"My biggest takeaway from my first day with Fable, low effort is the alpha."
Everyone runs Fable on high effort for everything. That's burning tokens on tasks a lower setting handles fine.
Greg's take: the honeymoon phase where Fable is bundled into subscription plans ends soon. Once you're paying per token, orchestration matters. He points at Factory's Droid as the tool that routes routine tasks to cheaper settings and saves the heavy mode for work that deserves it.
One tweet he shows even claims Fable on low beats Opus on high.
I run Claude agents every single day for my content pipeline, and this matches what I see. The real skill is knowing which 20% of your work actually needs the big brain, and running everything else cheap.
The landing page tournament
This is the prompt I'm stealing first.
Instead of asking Fable to "write my landing page" once, Greg runs a tournament:
→ Write the page 8 different ways, different hook and structure each time
→ Create 5 judges: a skeptical CFO, a distracted founder scrolling at midnight, a competitor, your ideal customer, and a conversion copywriter
→ Every judge scores every version and explains the score
→ Kill the losers, merge what worked, show the scoreboard
That's 8 versions, 5 judges, 40 scores, one winner.
He tested it on a backyard-home (ADU) checker idea from his idea browser. The judges came back scary good. The CFO docked points because a "$2,000 a month" claim reads like a Zillow estimate unless the page defends it immediately. The midnight founder flagged the line "$100,000 on a guess is a gut punch and it stops the scroll" as the thing that would actually make him stop.
Greg has written copy for products with millions of users, and he called the winning version incredible.
Why it works: one-shot output gives you one okay option. A tournament forces the model to argue with itself. Sub-agents holding distinct personas disagree, and disagreement is where good copy comes from.
The interview before the build
The second habit Greg wants you to steal: don't let Fable build anything until it has interviewed you.
The prompt: "I want you to build this rough idea. Don't build anything yet. Interview me like Mark Zuckerberg or Sam Altman or Brian Chesky. One question at a time, 15 questions max, hunting for the thing I haven't thought about. Push back when my answers are vague. After the interview, write the full spec and list three ways this fails, then build the V1."
He fed it a habit-tracker idea. Then the interview got uncomfortable.
When Greg answered a question with founder fluff, Fable replied: "That's half an answer."
When he said "I let life get in the way," it hit back: "That's a horoscope answer, Greg."
He said no LLM had ever talked to him like that, and he loved it. Because that's what a real operator would do.
The end result barely resembled the original idea. No widget. No habit platform. A phone number that reads your calendar, texts you at the right moment, and enforces a consequence you signed up for. The interview changed the product more than the build did.
Everyone one-prompting startup ideas gets pretty screenshots. The spec is where product-market fit hides.
Hire Fable to kill your company
Quick lightning round of the other use cases, because they stack:
→ The kill-your-company audit. Hand it your P&L, pricing page, churn data, and last 50 support tickets. Tell it: "You are a well-funded founder who hates me. Spend the day building the company that kills mine." Then, the twist that matters in 2026: rank every attack by how much of it the AI could execute itself with no employees.
→ The operating manual for working with you. Feed it 2 years of notes, decision docs, and postmortems (the 1M token context window holds years of you at once). Ask where you're systematically too early, too late, too optimistic, and what you always say right before a bad call.
→ The absence audit. Don't ask what's wrong with your business. Ask what's missing. "What is every successful company in my space doing that appears nowhere in my business? Rank by money left on the table."
→ The negotiation sparring partner. Make Fable become your counterparty, incentives and pressure included, and negotiate in rounds. After each round it breaks character and tells you what you just gave away for free.
→ The 80-page second opinion. Fable reads tables, footnotes, and schedules inside contracts now. Leases, term sheets, vendor agreements. Not a replacement for a lawyer, but it tells you where the flags are before you pay one.
→ Make it build its own tools. The most meta one: "Look back at everything I've asked you to do this month. Find the requests that repeat. Build yourself tools so next time it's one sentence instead of 10." Then ask what you're still doing by hand like an animal.
3 businesses you can start on top of this
Use cases are nice. Businesses are better. Greg closes with startup ideas that only became possible with Fable 5, and three stood out.
1. A synthetic focus group firm. D2C brands burn $50K in media spend learning an ad doesn't work. You run the tournament as a service: build 6 customer personas from 500 real customer reviews (actual review language, not demographics), then have every persona score 15 new ad concepts overnight. Output a kill list and a launch list. Charge $3,000 per launch against maybe $100 in tokens.
And yes, brands could do it themselves. Most won't. That gap is the business.
2. Custom software in 48 hours. Med spas, contractors, and clinics run on spreadsheets and group texts because custom software costs $80K a year. Sell internal tools at $5,000 flat, built in 2 days. The interview prompt IS your sales call: run it live on Zoom, the spec writes itself while the owner talks, Fable builds it after. Scoping was always the expensive part, and Fable now scopes big messes itself.
3. A contract refund firm. Mid-size companies sit on hundreds of vendor contracts nobody has read since signing. Auto-renewals, unused seats, prices ratcheting 8% a year. You cross-reference every contract against 12 months of invoices, rank the findings by dollars, and draft the exact renegotiation email for the top 10. You get paid 25% of savings found. Free unless you find money.
All three are the same shape as the ideas in my breakdown of Greg's agent arbitrage video: boring industry, painful spreadsheet, one model that suddenly does the expensive part.
My take: the window is the point
Greg says it straight at the end: we're in a honeymoon phase. Fable 5 is bundled into plans today, and that ends soon.
Which means right now is the cheapest R&D window you'll ever get. Every prompt in this video costs you nothing extra to test this week. In a few months you'll be paying API prices, and you'll want to already know which of these workflows earns its tokens.
I paid $13,000 for a mastermind when I was making $15K to $20K a month, because getting the playbook from people ahead of me was worth more than the cash. Six months later I was at $75K a month.
This video is that same trade, except free. Steal the prompts. Run the tournament. Let the model interview you before you build.
Just don't be in the 99% using the most powerful model ever released to write tweets.
FAQ
What is Claude Fable 5?
Fable 5 is Anthropic's newest flagship AI model, widely described as the most powerful coding and agent model available in 2026. It handles long-running autonomous work, reads full documents including tables and footnotes, and has a 1 million token context window that can hold years of your notes at once.
How do you make money with Fable 5?
Two routes. Use it inside your existing business (landing page tournaments, churn analysis, contract review, competitor audits), or build a service on top of it. Greg Isenberg's three favorites: a synthetic focus group firm at $3,000 per ad launch, $5,000 flat-fee custom software built in 48 hours, and a contract refund firm paid 25% of savings found.
What is a landing page tournament?
You have Fable write your landing page 8 different ways, then create 5 judge personas (a skeptical CFO, a distracted founder, a competitor, your ideal customer, a conversion copywriter) who each score every version and explain why. You kill the losers and merge the winners. It beats one-shot copy because the model argues with itself instead of agreeing with itself.
Is Fable 5 free to use?
For now it's included in many subscription plans, but Greg expects that honeymoon to end soon, with usage moving to API-style token pricing. His advice: learn the high-value use cases now while experimenting is cheap, and use orchestration tools to route routine tasks to lower effort settings.
Steal playbooks like this every week
This whole article came from one founder generously showing his prompts.
Every week on my podcast I sit down with bootstrapped founders doing $100K to $10M a year and pull apart exactly how their systems work: the real numbers, the real tools, the decisions behind them.
