You are using Claude Fable 5 wrong

In one sentence

To unlock the true power of Fable 5, you must move beyond "one-shotting" simple prompts and instead use it as an agentic engine that can simulate personas, mine massive datasets for hidden patterns, and aggressively push back on your vague ideas to reach Product Market Fit (PMF).

Why read this (and who it's for)

This is for founders, creators, and "vibecoders" who feel they are hitting a plateau with standard LLM interactions. It solves the problem of getting "okay" outputs by showing you how to architect complex workflows—like copywriting tournaments, founder interviews, and legal audits—that turn an AI model into a high-leverage business partner. After reading this, you will be able to build higher-quality copy, identify hidden revenue in contracts, and automate your own repetitive tasks using agentic loops.

The big idea

The central thesis is that most people are underutilizing Fable 5 because they treat it as a chatbot rather than an agentic engine. Because Fable 5 "explores before it answers," it is uniquely suited for high-reasoning tasks: orchestrating complex multi-step workflows (like video editing), acting as a pro-negotiator, or mining years of personal data to find systematic flaws in your decision-making.

The logic clicks when you realize that the quality of the output is directly tied to the rigor of the process. Instead of asking for "a landing page," you should be running a "tournament" where the AI pits different versions against each../../ judges. Instead of asking for "an app idea," you should be undergoing an interview with a simulated Mark Zuckerberg to find the flaws in your logic before a single line of code is written.

Key ideas

Agentic Orchestration

Fable 5 can act as a director, not just a writer. It can manage sub-agents to perform specific tasks (like transcription, JSON creation, and ffmpeg execution) to achieve a complex final result like professional video editing. Why it matters — It allows you to automate high-skill production workflows that previously required hundreds of thousands of dollars or dozens of manual hours. In the book — The author describes an Anthropic employee who used Fable to edit a launch video by orchestrating 11 Labs, JSON files, and ffmpeg to stitch together shots and color grade them perfectly.

The Content Brain

AI content often feels "not that good" because you haven't provided the right strategy or inputs. To make it work, you must build a "Content Brain"—a repository of your origin story, known-for, offer, ICP (Ideal Customer Profile), and frameworks. Why it matters — It ensures the AI isn't guessing your voice; it is executing on your specific brand pillars. In the book — The author suggests using a tool like Obsidian to house these pillars so Fable can run a weekly loop of research, niche scanning, and hypothesis testing.

Low Effort is Alpha

Not every task requires the most powerful model. Using "High" settings for routine tasks is an inefficient use of tokens and resources. Why it matters — It optimizes your costs and allows you to route simpler tasks to smaller models (like Opus) while reserving Fable 5 for high-reasoning "Heavy" lifts. In the book — The author mentions using products like Droid from Factory.ai to orchestrate when to use Fable vs. lower-effort models.

Tournament Frameworks

Instead of asking for one option that might be "okay," you should pit multiple versions against each other using simulated judges to find the winner. Why it matters — It forces the model to critique itself, leading to a final version that is significantly higher in quality than a single-shot prompt. In the book — The author uses a "Landing Page Tournament" where 8 different versions of copy were judged by 5 personas (a skeptical CFO, a distracted founder, etc.) to produce a winning verdict.

The Interview before the Build

Most people one-prompt startup ideas and end up with visually great products that have low likelihood of PMF. You must use F Why it matters — It forces you to move past "horoscope answers" (like "life got in the way") and identifies the actual friction points in your habit or business model. In the book — The author used a prompt to simulate a founder interview with Brian Chesky/Sam Altman, who pushed back on his vague answers until they arrived at a spec for a phone-number-based habit app with consequences.

Absence Mining

Most people ask AI what is wrong with what exists. You should instead point it at your analytics and content calendar and ask what is absent. Why it matters — It identifies the revenue lines or customer segments you aren't even trying to reach, which is where the real growth lies. In the book — The author describes pointing Fable at his own business data to find what successful competitors are doing that he is completely ignoring.

Decision Mapping

With a 1-million token context window, you can feed the AI years of your personal notes, emails, and postmortems to find patterns in your own behavior. Why it matters — It reveals systematic errors (being too early, too late, or too optimistic) that you are too "inside" the business to see yourself. In the book — The author suggests creating an "operating manual for working [him]" based on his past decision docs and postmortems.

Frameworks & tools

The Landing Page Tournament

Use this when you need high-converting copy for a new offer.

1. Generate: Ask Fable to write your landing page in 8 different ways (different hooks, structures).
2. Judge: Create 5 judges (e.g., Skeptical CFO, Distracted Founder, Competitor, Ideal Customer, Conversion Copywriter).
3. Score: Have every judge score every version and explain the scores.
4. Synthesize: Kill the losers, merge what worked into a final version, and show the scoreboard.

The "Interview before the Build" Prompt

Use this to move from a vague idea to a high-likelihood PMF spec.

1. Persona: Tell Fable to act as an expert (e.g., Mark Zuckerberg or Brian Chesky).
2. Constraints: Limit it to 15 questions max, one at a time.
3. Pushback: Explicitly tell it to push back on vague answers and hunt for the thing you haven't thought about.
4. Output: After the interview, demand a full spec, three ways the idea fails, and the V1 architecture.

The Contract Refund Audit

Use this to find hidden money in vendor contracts.

1. Input: Upload all vendor contracts and 12 months of invoices.
2. Cross-reference: Ask it to find auto-renewals in the next 90 days, mismatched rates, and unused seats.
3. Rank: Rank findings by dollar amount.
4. Action: Draft exact emails to vendors including the specific leverage you have.

Memorable examples & stories

The Video Edit: An Anthropic employee used Fable to orchestrate a professional video edit from 17 takes, using ffmpeg and JSON files to color grade and stitch shots perfectly. Lesson: Use Fable as an agentic director for complex production workflows.

The Backyard Home Copy: The author took a vague idea from Ideabrowser, put it through the "Landing Page Tournament," and got copy so good that a "skeptical CFO" judged it as a winner because it defended its claims immediately. Lesson: Tournaments produce higher-quality copy than single prompts.

The Habit Widget Interview: When asked about why he failed his workout streaks, Fable pushed back on his "life got in the way" answer, forcing him to realize that timing was the problem. Lesson: Use AI to find the truth, not just to agree with you.

The Company Killer: The author gave Fable his P&L, pricing page, and support tickets and asked it to build a company to destroy his own. Lesson: Use AI to find your vulnerabilities by flipping the perspective.

Key quotes

"Low effort is the alpha." Why it matters: Don't waste high-reasoning tokens on routine tasks; orchestrate smaller models for the "easy" stuff.

"Timing is the problem. A smarter ping fixes it." Why it matters: This was the core insight discovered during the founder interview regarding habit apps.

"Don't tell me what's wrong with what exists. Spend your time on what's absent." Why it matters: Growth comes from identifying the revenue lines and customer segments you are currently ignoring.

How to apply it

This week

• Run a Copy Tournament: Take one landing page or email you are struggling with and run it through 8 versions and 5 judges.
• The "Absence" Audit: Point Fable at your current marketing funnel/analytics and ask: "What is every successful company in my space doing that appears nowhere in my business?"
• Contract Scan: Upload one vendor contract and ask for a list of auto-renewals or price escalators.

Ongoing

• Build your Content Brain: Start a dedicated file/folder of your origin story, ICP, and brand pillars to use as a permanent input.
• The Self-Automation Loop: Every month, ask Fable: "Look back at everything I've asked you to do this month. Find the requests that repeat. Build yourself tools and reusable instructions for each one."

Common pitfalls

• One-prompting: Asking for a final product in one go without an iterative process (like a tournament or interview).
• Accepting "Horoscope" Answers: Letting the AI agree with your vague excuses instead of forcing it to push back on your logic.
• Ignoring the Context Window: Not feeding the model enough data (years of notes, full contracts, etc.) to find the actual patterns.

Connects to

• Agentic Workflows: Extends the idea of AI agents by providing specific "managerial" prompts for video and copy.
• Product Market Fit (PMF): Challenges the standard "build then measure" loop by suggesting an "interview before the build" phase.

If you remember only three things

• Agentic Pushback: Use Fable to interview you and find flaws in your logic before building.
• Tournament Frameworks: Pit multiple versions against simulated judges to reach higher quality.
• Absence Mining: Focus on what is missing from your business, not just fixing what exists.

Internalize it

• Reflection questions

  • Which of my current business ideas am I "one-prompting" without actually vetting for PMF?
  • If I gave Fable all my past decision docs, what systematic pattern would it find in my failures?
  • What is the one thing a competitor would do to kill my company today that I am currently ignoring?

• Flashcards Q: What is the "Content Brain"? A: A repository of origin story, known-for, offer, ICP, and frameworks used as permanent inputs for AI content. Q: Why use a "Tournament" framework for copy? A: It forces the model to critique itself against multiple personas, leading to higher quality than single-shot prompts. Q: What is "Absence Mining"? A: Pointing an AI at your data and asking what successful competitors are doing that you are NOT doing. Q: How do you use Fable for video editing? A: By using it as an agentic director to orchestrate transcription, JSON files, and ffmpeg execution. Q: What is the "Interview before the Build" prompt? A: A prompt where Fable acts as a founder expert (e.g., Zuckerberg) to push back on your vague ideas until a high-likelihood PMF spec is reached.

Fidelity note

Confidence: High. This summary captures the specific prompts, startup ideas, and tactical workflows provided in the transcript. I am certain about the "Tournament," "Interview," and "Absence" frameworks. Note that some technical specifics (like exact API pricing for Fable) are mentioned as future-looking/speculative by the author.