Big Idea

The argument in one line.

The people extracting outsized value from Fable 5 are not using it as a chat tool — they are pointing its million-token context window at the datasets most people ignore and letting sub-agents push back, argue, and build reusable tools on their behalf.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

A solo or small-team builder who wants concrete, copy-paste prompts they can run with Fable 5 today.
Someone with unused business datasets — churn reports, support tickets, vendor contracts, years of personal notes — who wants to know how to extract value from them.
A founder or freelancer looking for startup ideas that are structurally only viable now, with exact prompts and business models included.
Anyone concerned about token cost as Fable 5 moves toward API-only pricing and wanting a framework for effort-level management.

SKIP IF…

You want an architectural deep dive on how Fable 5 differs from previous Claude models at the parameter level.
You are not using the Anthropic Claude suite — the prompts and configuration advice are Claude-specific throughout.
You need advanced implementation tutorials; this is use-case and prompt strategy, not technical coding instruction.

TL;DR

The full version, fast.

Fable 5's real edge is not its benchmark scores but the combination of a million-token context window, strong visual reasoning, and sub-agent orchestration that lets it hold years of your data and refuse vague input. The video covers eight prompts: copywriting tournaments scored by AI judge panels, an interview-before-build workflow that role-plays founder-level pushback, company red-teaming via P&L and churn data, decision archaeology on personal notes, gap detection ranked by money left on the table, negotiation simulation, contract analysis, and self-building tool creation. Three startup ideas close the video: a synthetic focus group firm charging 3k per launch, 48-hour custom software at 5k flat, and a contract refund firm taking 25% of savings found.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 02:22

01 · Intro

Sets up the premise: move past benchmarks, get tactical on use cases and startup ideas.

02:22 – 05:50

02 · Anthropic Employee Edits a Launch Video With Fable

End-to-end AI video pipeline: 17 takes, 4 scenes, sub-agents pick best shots, FFmpeg stitches, color grading from scratch, Remotion design frames, Figma MCP export.

05:50 – 07:30

03 · Building an AI Content Engine

The 4-System AI Content Engine: inputs (origin story, ICP, offer, frameworks) into content brain into weekly research/performance loop. Fable runs everything and reads your charts.

07:30 – 08:42

04 · Best Way to Configure Fable 5

Low effort is the alpha. Fable Low beats Opus High on benchmarks. Use factory.ai Droid to orchestrate effort levels and keep token spend lean.

08:42 – 13:18

05 · Prompt 1: Copywriting Tournament

Write 8 landing page versions, create 5 AI judge personas (CFO, midnight founder, competitor, ICP, conversion copywriter), 40 scores, one winner. Live demo on a backyard ADU SaaS idea.

13:18 – 18:34

06 · Prompt 2: Interview Before the Build

Role-play as Zuckerberg/Altman/Chesky. 15 questions max. Explicit pushback on vague answers. Full spec + 3 failure modes + build v1. Demo builds a habit-accountability SMS app.

18:34 – 20:18

07 · Prompt 3: Hire Fable to Kill Your Company

Feed P&L, pricing page, churn data, last 50 support tickets. Fable plays a funded founder who hates you. Ranks every attack by how much you could execute with no employees.

20:18 – 21:20

08 · Prompt 4: Your One-Page Operating Manual

Feed years of notes, decision docs, postmortems. Find decision patterns and systematic biases. Output: the operating manual for working with you that a COO would secretly keep.

21:20 – 22:06

09 · Prompt 5: Find the Gaps Worth Filling

Point at analytics, content calendar, and offer. Ask only what's absent — not what's wrong with what exists. Rank by money left on the table.

22:06 – 23:11

10 · Prompt 6: Negotiation Simulator

Fable becomes the counterparty with their incentives, alternatives, and pressure. Negotiate in rounds. Breaks character to debrief what you gave away for free.

23:11 – 24:56

11 · Prompt 7: The 80-Page Second Opinion

Feed the full contract plus all exhibits. Get three lists: hidden costs, 18-month regrets, missing protections. Plus exact ask phrasing for the top 3 changes.

24:56 – 25:47

12 · Prompt 8: Make Fable Build Its Own Tools

After weeks of use, ask Fable to study your request history, find repeating patterns, and build reusable one-sentence tools for each. The most meta use case.

25:47 – 31:23

13 · Startup Ideas Only Possible Now

Three ideas: synthetic focus group firm (3k per launch, 100 in tokens), 48-hour custom software (5k flat, interview prompt is the sales call), contract refund firm (25% of savings found, free otherwise).

31:23 – 33:08

14 · Closing Thoughts

Use the honeymoon window while Fable 5 is included in plans. Get fluent now before API pricing changes the economics.

Atomic Insights

Lines worth screenshotting.

Fable Low beats Opus High on routine coding benchmarks — defaulting to the highest effort setting is just burning tokens with no upside.
Sub-agents can hold distinct personas and argue with each other instead of a single model agreeing with itself — that structural difference is what makes AI judge panels produce real variance in copy.
The million-token context window can hold years of notes, emails, and postmortems simultaneously, which is what makes decision archaeology on your own patterns actually work.
DTC brands spend 50k on media before learning an ad fails; running a 50-variant AI judge tournament costs roughly 100 in tokens.
Scoping was always the expensive part of custom software — Fable now interviews the client, writes the spec while they talk, and builds after. That changes who can sell custom software.
Contract auto-renewals, unused seats, and 8% annual price escalators hide in tables and footnotes — exactly where Fable improved document reading makes it valuable.
When a model is told to role-play as a founder hunting for PMF it will say that is a horoscope answer instead of validating vague input — and that pushback is what produces a spec worth building.
The most meta use case: after a few weeks of use, ask Fable to study your repeated requests and build reusable one-sentence tools for each pattern.
Hiring AI to red-team your own company ranks threats by which ones could execute themselves — the only threat ranking that matters when anyone with Fable 5 can spin up a competitor.
Fable 5 is currently included in the Max plan; when it moves to API-only pricing, the economics of every use case in this video change materially, so the window to learn is now.

Takeaway

Eight prompts that treat Fable 5 as a system, not a chat box.

WHAT TO LEARN

The gap between average and outsized results with Fable 5 comes down to whether you give it a single question or a dataset, a vague idea or a structured interview, one shot or a tournament.

Running a copywriting tournament — eight versions scored by five distinct AI judge personas — produces copy with real variance instead of a single polished guess that all feels the same.
The interview-before-build prompt works because it explicitly role-plays someone who wants the product to succeed, not someone who wants to please the prompter; the pushback is the feature.
Pointing Fable at churn data, support tickets, and a P&L produces a competitive threat map ranked by which threats can self-execute — more honest than asking for generic growth ideas.
The million-token context window changes what feeding Fable your history means: years of notes, emails, and postmortems fit in one session, surfacing decision patterns you cannot see from inside.
Asking what is absent from your business — not what is wrong with it — is a structurally different question that surfaces revenue lines that should exist but do not.
Fable can simulate a counterparty in a negotiation with their actual incentives, pressure, and alternatives, then break character to debrief what you conceded — that is practice, not just advice.
Contract value hides in tables and schedules: auto-renewals, unused seats, and price escalators that ratchet 8% a year. Those sections are now readable at scale for the first time.
After a few weeks of use, asking Fable to study your own request patterns and build reusable tools is the compounding move — each tool reduces the next task from ten sentences to one.
Low effort is the correct default; Fable Low outperforms Opus High on routine work, and reserving Max effort for work that directly earns money is what keeps token costs from exploding once API pricing lands.
The three startup ideas share a structural logic: each is an existing service category where the cost of delivery just dropped by 90%, creating a window to undercut incumbents at high margin.

Glossary

Terms worth knowing.

Copy tournament: A prompt workflow that generates multiple versions of the same copy, assigns distinct AI judge personas to score each version, kills the losers, and merges what worked — producing genuine variance instead of one AI guess.
Interview-before-build: A prompt pattern where the AI role-plays as a seasoned founder and asks up to 15 pointed questions before writing any spec, explicitly pushing back on vague answers to improve product-market-fit odds.
Effort orchestration: Routing different tasks to different effort levels (Low / Medium / High) in Fable 5 to minimize token cost, using tools like factory.ai Droid to automate the routing decisions.
Decision archaeology: Feeding years of personal notes, emails, and postmortems into a large-context model to surface patterns in past decisions — where you are systematically too early, too late, or too optimistic.
Sub-agent: An autonomous AI process that operates within a larger workflow, able to hold a distinct persona, execute tool calls, and return results without human prompting at each step.
ADU: Accessory Dwelling Unit — a secondary residential structure built on the same lot as a primary home, commonly called a backyard home or in-law suite.
PMF: Product-Market Fit — the degree to which a product satisfies a strong market demand, typically measured by retention, word-of-mouth growth, and willingness to pay.

Resources

Things they pointed at.

00:00productFable 5 Prompt Pack ↗

07:58toolfactory.ai Droid ↗

09:25toolideabrowser.com ↗

04:14toolElevenLabs ↗

05:00toolRemotion ↗

05:29toolFigma MCP

06:47toolObsidian ↗

Quotables

Lines you could clip.

16:16

“That's a horoscope answer, Greg. Life got in the way is what everyone says when they haven't watched the tape.”

Fable roasting the user on vague input — visceral and specific→ TikTok hook↗ Tweet quote

07:31

“Low effort is the alpha. Incredibly smart, and cheaper than Opus — what is your rush?”

Counterintuitive claim with a clean one-liner format→ IG reel cold open↗ Tweet quote

25:34

“What should I be delegating to you that I am still doing by hand like an animal?”

Memorable, repeatable prompt framing — quotable out of context→ Newsletter pull-quote↗ Tweet quote

19:24

“I would rather hear it from you than read it in my churn dashboard.”

Strong punchline for the kill-your-company prompt→ Newsletter pull-quote↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

metaphoranalogy

00:00I've got a dumb question. If the world's most powerful AI model just came out, Fable five, how are we supposed to use it to make money and be more productive?

00:09I mean, I've seen some crazy examples on x of people one shotting like a Monopoly game, and that's really cool. But how could I use this super intelligence and just point it to different problems to make me money and be more productive? That's what this episode is about.

00:25By the end of this episode, you're gonna understand why Fable five really matters. But more importantly than that, you're gonna understand practically how you can use Fable five to make money and be more productive.

00:38Now is the time that you have this unfair advantage for actually pointing this thing, this insanely powerful model at problems that could build businesses, that could make you more productive.

00:51And to be honest, it's a lot of fun. So enjoy the episode, and I'll see you in there.

01:04Oh, I needed to make this video. I needed to make this video because 99% of people are using Fable five wrong.

01:11I mean, this is the most powerful model that we've ever seen, and people are just they're just using it the wrong way.

01:18They're not seeing all the opportunities, all the ways to make money, be more productive with it. So I had to make this episode where I break down how you're using Fable five wrong, what are a bunch of use cases and prompts that you can use today that's gonna help you, and a bunch of startup ideas that now with the most powerful coding agent ever available to mankind, what has that unlocked?

01:43That's what this episode is all about. This is a huge unlock, yet all the episodes that I've seen and tutorials of videos that I've seen on the Internet have been about benchmarks and things like that.

01:56But I wanted to get to the tactical use cases for for Fable five, so that you, after this video, if you stick around to the end, you understand how to use Fable five in terms of use cases and in terms of startup ideas, so that you can actually build a business and start making money and with Fable five.

02:18Let's get right into it. The first thing is this guy who works at Anthropic used Fable to edit and launch his own launch video, and it looks spectacular.

02:33I can't believe how good it is. And I will say, he definitely had a good camera.

02:42You can see here, he, you know, he's got a good setup. He's got this cool bookshelf.

02:48But he did 17 takes and four scenes, and it ended up with this beautiful visual.

02:55Right? And you can see the text looks amazing. This is something that would have cost hundreds of thousands of dollars to make, and using Fable with two video edit is totally possible.

03:05The prompt is here, and I'm gonna read it to you. I'm processing the recording of a bunch of videos in Fable full recording. The script for them is in Fable full recording, Fable five script.

03:16I want you to run the Eleven Labs transcription service on them and stitch together the best shots into one final clip. A few notes, there are multiple takes per scene. The best takes are usually the one at the end with the fewest ums, but that can can change.

03:31I also reshot the first scene at the end. They should all be organized by timeline. For a few of them, I start the video by saying, hey name, as a way of starting the sentence, warm, we should cut that out.

03:42Create a JSON file, use f m m p g, orchestrate this using workflows slash gold, don't stop until you have the final video.

03:51So what do we learn from here? We know that he used 11 labs, he stitched together the best shots, they created a JSON file, he used FFmpeg, and he orchestrated all of this using workflows using slash goal, and said, don't stop until you have a final video.

04:08And and it worked, you know. I can go through this deck, but you can see here, he uses whisper whisper whisper flow, it removed all the ums, the sub agents picked the best takes, the edit is in this JSON file, and then FFmpeg executes the JSON.

04:28It stitches together the first cut from the prompt, create a final scene using FFmpeg. It actually color grades with this prompt.

04:40The color grading feels a bit muted. Can you fix that? It's able to color grade, and it looks amazing.

04:46Look at that. Just absolutely perfection. That's beautiful.

04:50And anyone who's ever done any color grading before knows how hard this is.

04:57The the static design frames are all there.

05:02He uses Remotion to stitch it all together. Each PNG is rebuilt as code. And then there's one file that sets the field, so you got this, you know, beautiful beautiful one file here.

05:20And the point here is you can create a beautiful a darn beautiful video.

05:26You can export it to Figma. You can use the Figma MCP, and you can just use Fable to build something beautiful.

05:34You know? So many people so many of us are vibe coding apps, can't get people to those apps.

05:40But when you have something like Fable now, you can use it to actually grow your whatever it is you're working on.

05:47This is a great example of that. That's one example. The two is, and this is, uh, I saw a wizardecom tweeted this.

05:56You can create a purse you can create a an AI content engine. I saw other people basically asking Fable to create tweets, and they were saying, uh, it's not actually not that good at writing.

06:07It's not really good at writing. It's not good at writing because you haven't put the right inputs, the right strategy, the right weekly engine, and then connect it to Fable.

06:18So if you wanna use Fable as a content engine, here's how you can do it. You're gonna want to put your inputs, so you're gonna want your origin story, you're known for, your offer, your ICP, your frameworks, and you can ask Fable to actually help you with that.

06:35You can put that in your content brain. You can use something like Obsidian here, where your pillars, your funnel structure, your tone. Then every week, you can do this, you know, research, scanning niches, finding topics, performance, uh, three hypothesis test weekly, and that creates the assets as this weekly loop and that uses Fable five.

06:56So Fable five runs every system. It reads your charts. The great thing one of the great things about Fable five is it's able to visually check out using Cowork, you know, browser using Cowork.

07:11It's able to go and and and see things and really, really understand it, and it works for hours on its own. And the third piece of of thing that I think people are getting wrong with Fable, and then I want to get into actually going into into Claude, showing some use cases, and giving startup ideas, is my friend Morgan Linton said this.

07:33He says, my biggest takeaway from my first day with Fable, low effort is the alpha. So the mistake a lot of people are making is, you know, the default's on high. But I think that as you know, I think it's June 22, you're it's gonna be API pricing.

07:48It's not gonna be included in your Max account. You're not you know, you're gonna wanna obviously not spend a ton of tokens.

07:54So how do you do that? You're gonna wanna use something like factory.ai. They have a product called Droid, and Droid is gonna basically orchestrate when you're using Fable, what effort you're using at, and when are you using something like Opus or other models that are more routine tasks.

08:11So I think you're gonna start to see, uh, just more people using things like Droid, um, more people actually doing lower effort for more routine tasks.

08:21And just so you know, you know, someone tweeted this here. In fact, Fable low is better than Opus high, so it's pretty impressive. So if people want me to do a full breakdown on how to use Droid and Factory AI to set up this orchestration, just comment.

08:38I can do it, and, uh, I'll show you how you can set that up. Alright. Let's talk about use cases.

08:45Let's start with this one, actually. If you're anything like me, writing copy is really difficult.

08:52And one thing I'm learning is that, you know, copy is you know, one thing I'm actually relearning is that if you have good copy, you're going to be able to your conversion rate's gonna be higher.

09:05So what's really cool is you can use you can actually set up a copywriting tournament so that you end up creating absolutely stunning websites, building websites using Fable with incredible copy.

09:23So look at this one prompt that I got for this SaaS idea.

09:29I took the SaaS idea from ideabrowser.com. I screenshot it, I put it in, and I was able to get this incredible idea around backyard homes.

09:38So you put in your, um, your address, it checks your lot, and says, hey, can you build, uh, these ADUs, which are basically mini houses in your backyard?

09:49And when you look at this copy, can you build a backyard home on your lot? Enter your address. In minutes, you'll have a written verdict.

09:57This is a real report. Yours looks like this, and it actually creates it. Look how beautiful this is.

10:03The little animations. The copy is incredible. I'm going to show you how to do this.

10:09And the way I do it is with with something called the landing page tournament. So let's go into the into the prompt. By the way, in this episode, I'll include all prompts in in the description, and so you can go and and and grab them.

10:24So I say, write my landing page eight different ways, different hook, different structure each time, then create five judges, a skeptical CFO, a distracted founder scrolling at midnight, a competitor, my ideal customer, and a conversion copywriter.

10:40Have every judge score every version and explain the scores. Kill the losers, merge what worked into a final version, and show me the scoreboard so I can see why it won. And then I just basically attached the idea of the day, can I build a backyard home tool to find in minutes from ideabrowser.com?

10:58I put that in there just to see if it would work. And let's actually look at, uh, what happened. So what's great about doing these tournaments, so I use Fable five high over here, is it basically, uh, instead of getting one option that maybe is okay, you're basically getting it you're pitting it against each other, and you're getting just a lot a lot better a lot better output.

11:24So when we go to the landing page tournament MD, you we basically got eight versions of copy. We had five judges, 40 scores, and one winner.

11:33And it just does such a good job. I'll go through one of these, and and then I'll go to the end just so you can get a sense. So actually, I'm just gonna skip to the I'm gonna skip to the end here just so you can see what happens.

11:45So it created a scoreboard final version. So the CFO scored at this, the founder scored at this.

11:54The competitor scored at that. And then based on that, it said, this is, you know, just a lot better.

12:02So I have been writing copy for a lot of years on on websites and apps.

12:10I get, you know, millions I have millions of users. And when I saw the backyard copy, like the one that won, I I I was like, this is incredible.

12:20And the the secret sauce is that I use this I use this framework.

12:29You can see here, like, the CFO says, yeah. I'm gonna score this a six because 2,000 a month is a claim the page must immediately defend, or it reads like a Zillow estimate. This is what's so crazy about Fable five is it's it's scary it's scary good.

12:46Like, yeah, a CFO might say that. The the midnight founder who's scrolling in the middle of night reads, a $100,000 on a guess is a gut punch, and it stops the scroll.

12:58You know? It just it's so good, and this is just a really it's something that I'm seeing no one no one use that I think you should be using as well.

13:12Another thing that Fable five is just incredible at is planning. This is something that if you're building a startup, it's worth considering.

13:22So the interview before the build. So the prompt goes, I want you to build this rough idea.

13:30Don't build anything yet. Interview me like Mark Zuckerberg or Sam Altman or Brian Chesky, experts at building companies that get PMF.

13:39One 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, and build the v one.

13:51So what I'm seeing on x and other places is people are just one prompting these startup ideas, and they're not getting really what they want.

14:03It might look visually great, but from my perspective, I don't think it's something that has a high likelihood of PMF, product market fit. So we're actually gonna use this prompt to let me go scroll up.

14:18We're gonna use this prompt so that we can have a spec, a full spec that has a highest likelihood of success. So I grabbed a idea of the day from ideabrowser.com.

14:31It's basically this habit widget that pings you at the right moment. Imagine a habit app that basically gives you a push notification and says, hey, you need to work out or you need to drink water or something like that. And what's so crazy about, uh, Fable five high in this case is it asked me questions.

14:51We're gonna go through some of the questions. It says, what habit have you personally failed to keep in the last twelve months? And I said, failed workout treats.

15:00And then I said, and I'm just a person who wants to unlock my potential and can always do better. And then it goes, that's half an answer. Like, I'm getting pushback from Fable.

15:15Fable feels like a real person. If that was a real person, if that was a Brian Chesky, if that was a Sam Altman, and they were actually trying to help me build an app that gets the PMF, they would say something like that.

15:27And to this date, I have never had an LLM speak to me like that, and I absolutely loved it. Because that's what I need in order to create an app that has a high likelihood high likelihood of success.

15:43So it says, walk me through the actual death of your last workout streak. What were you supposed to work out? What time of day?

15:49What happened on the day you missed? Did you miss because you forgot or because you remembered and chose not to? This matters more than anything in that screenshot.

15:58Uh, the entire thesis is timing is the problem, a smarter ping fixes it. And then I said, I let life get in the way. Then it goes, that's a horoscope answer, Greg.

16:10Life got in the way is what everyone says when they haven't watched the tape. You never let a podcast guest get away with that, and you'd say, give me the scene.

16:22You love to see it. It pushes back and it helps me you know, I'm not gonna go through all of this because that's not the point. The point is it pushes back on on you, and by the end of this by the end of this, we got an incredible, incredible outcome.

16:38And I wanna I wanna talk about it. It says, the interview changed the product more than you might realize. What we're building shares a name with the Idea Browser write up and almost nothing else.

16:47No widget. Widgets can't ping you. No week of manual check ins.

16:53You banned homework. No habit platform workouts only. It's a phone number that can see your calendar and text you with receipts and has a consequence consequence you signed up for.

17:03So basically, what it did is it took this idea of like, you're gonna get a text message. It well, you're gonna connect it to your calendar, and it's gonna basically say, you should work out now. And if you don't, you're gonna get a a consequence.

17:16It built the app. Completely works. It built the spec document, and and and it's just something that has a high likelihood of success.

17:30So I don't see anyone doing this using Fable five, helping you interview.

17:37You you know, the key is say say it's a Mark Zuckerberg, say it's a Bell Brian a Chesky, and then you get this spec document that really just figures out who you are, what the onboarding should be, what the loop should be, and just gives you the highest likelihood of success and gives you the architecture.

17:56You can see here it says build with Python, fast API, AP schedule scheduler, what the pricing should be, and what the scope which should be out of which should be out of scope for v one.

18:09So just an incredible output, and I I've never seen an LM do something so high quality.

18:18We're going to we're gonna go through a few more use cases. I'm just gonna just ex just lightning round it lightning round it.

18:25I'm not gonna show it in in in Claude, but it's just to help you get your creative juices flowing for, okay, how else can I use this thing? Well, you can hire it to kill your company.

18:38So everyone asks for AI everyone asks AI for growth ideas, but you can flip it. Give it everything and a day to destroy you.

18:46The magic is it ranks threats by which one could execute itself, which is the only threat ranking that matters now.

18:53So here, check this out. You say, here's my p and l. Here's my pricing page, my churn data, my last 50 support tickets.

19:01You are a well funded founder who hates me. Spend the day building the company that kills mine. Real plan, positioning, pricing, first 10 customers you'd steal, and the exact email you'd send to each one.

19:13Then rank every attack by how much of it you can do yourself with no employees. Don't be polite. I'd rather hear it from you than read it in my mind in the churn dashboard.

19:24This is the output of this is incredible. So the insight here is you wanna mine your support tickets.

19:32Fable is incredible with big datasets. It can look at your churn data churn data, and it's going to help you make more money from a SaaS. Now you can also go and buy a SaaS now.

19:46You can actually go and refactor that code using Fable. You can go and look at those support tickets. I've talked about this on the channel.

19:52There's an opportunity, especially now with Fable five, to actually go buy things for not a lot of money, kind of flip it, you know, or renovate it, maybe flip it, or just keep it for profit and cash flow.

20:06But the idea is basically now, uh, you have something that is just so high quality. Um, we talked about the copy tournament. Um, this is a is a really interesting one.

20:17So you give it two years of your notes, your decision docs, your postmortems. You ask it to map every meaningful decision.

20:26What I chose, what I believed at the time, what actually happened, find my patterns, Where am I systematically too early, too late, too optimistic? What do I always say right before a bad call? Write me the one page, operating manual for working Greg that a COO would secretly kept keep.

20:46So what's really cool about Fable five and where we are today, 2026, is you've got a 1,000,000 token context window, which can hold years of you at once.

20:57And I don't see that many people doing this. So maybe you don't have old notes, but you definitely have emails, you definitely have text messages, and ask it to find patterns in your decisions that you can't see because you're in the inside.

21:12And then just set up routines that remind you of things, right, and make you a better person. That's something that I think we're gonna see more of.

21:20Number five. So Fable's really good at noticing what's missing because it explores before it answers. So you can actually point that at anything, like your calendar, or your code base, or your marketing funnel.

21:35So you can say, here's my analytics, my content calendar, and my offer. Don't tell me what's wrong with what exists.

21:42Spend your time on what's absent. What is every successful company in my space doing that appears nowhere in my business? What customer am I not even trying to reach?

21:53What revenue line should exist and doesn't? Ranked by money left on the table. So no brainer, and something that I started doing this morning, and I have gotten some really interesting insights.

22:07Negotiation. You know, if you're in business, you're gonna be doing negotiations.

22:13And a lot of people actually go to, like, Perplexity and those or or ChatGPT, and they'll say, like, hey, I'm gonna respond to this email.

22:23How would you respond? Now with Fable five, you can actually actually have, like, basically a pro negotiator at the table.

22:30Here's what you say. I'm negotiating a deal with this counterparty. Here's what I know about them and what I want.

22:36Become them. Their incentives, their alternatives, their pressure.

22:40We negotiate in rounds. You respond the way they actually would, including going quiet or getting aggressive.

22:48After each round, break character and tell me what I just gave away for free. Don't let me win because I'm the user. So just, you know, easy way to to to to just win more deals, get get more out of them, and here's the prompt.

23:07We're gonna do two more use cases, and we're gonna get into startup ideas. So you can get a 80 page second opinion with Fable five.

23:18So because Fable five actually reads tables now, footnotes, charts inside documents, it can see it, um, that's where the bodies are buried. Right?

23:27Leases, term sheets, insurance policies, vendor contracts, people are seeing incredible, uh, I think it was Aaron Levy from Box actually said that, uh, just in terms of legal, you know, he he they've been seeing just incredible outputs from Fable five.

23:47So here's here's the prompt, here's the full contract, every exhibit included, read all of it, especially the tables and schedules, three lists, what cost me money that isn't obvious, what I'd rather get in eighteen months, and what's missing that should be here to protect me.

24:04Then tell me the three changes you'd request and exactly how to phrase the ask. Flag anything when I should pay where I should pay a real lawyer to look. Now I'm not suggesting that you obviously just use AI to to do your legal work, but let's be real.

24:22Lawyers cost a lot of money. And if you're in a startup, every dollar counts. And some things, you know, I'm just speaking for myself.

24:31Some things, I'm like, I I don't need to send I don't need to send this to a lawyer. And some things, I just want to know where the flags are.

24:39So no matter if you're working with a lawyer or not, you know, do your own research. I'm not saying that you shouldn't work with a real lawyer.

24:46I'm saying that even if you're working with a real lawyer, you should put it through Fable five and say, what are the flags, and here's the prompt that you're gonna need for that. Boom.

24:57Eight, make it build its own tools. This is the most meta one, and it's the one that actually separates people who get it. So after you've used it after you've used Fable five for a few weeks, just ask it to study the work and automate itself.

25:11Here's here's what I mean. You prompt it. You say, look back at everything I've asked you to do this month.

25:17Find the request that repeat. Build yourself tools and reusable instructions for each one, so next time it's one sentence instead of 10. Then tell me, based on what I keep asking, what should I be delegating to you that I'm still doing by hand like an animal?

25:35And you'll be shocked. I've even done this just over the last twenty four hours. What you've been able to automate and and what you're able to and what kind of software you're able to create here.

25:47Alright. I'm gonna go over a few startup ideas that are only possible now with Fable five. I have five.

25:56I'll include the full list and the full prompts in the in the description. I'm only gonna have time to go over three. So we're gonna go over my three favorite ones.

26:06The first is a synthetic focus group firm. So d to c bat brands will spend like $50,000 on media before learning an ad doesn't work.

26:16So remember we talked about the copy tournament? Well, you run the tournament as a business. 50 ad variants judged overnight by panels built from the brand's actual customer reviews, and then you charge something like 3,000 per launch, and maybe that costs you a $100 of tokens.

26:33And the brand's media buyer becomes your repeat customer. So this is only possible now because sub agents can hold distinct personas and argue instead of a model agreeing with itself.

26:45Okay. Here's what I mean. Here's the prompt.

26:47And this is by the way, this is just gonna be helpful for anyone even if they don't wanna use the start up idea, but if they're doing any paid ads themselves, this is gonna be helpful, super helpful.

26:58So you give it 500 customer views. Here are 500 customer views for athletic greens and their 10 best ads.

27:06Build six customer personas from actual review language, not demographics. Then take these 50 new ad concepts and have every persona score every ad.

27:17Would they stop scrolling? Would they click? What's their objection?

27:21Output, a kill list, a launch list, and the one insight about this customer that the brand that the brand doesn't know. So someone in the comments section, I can hear them I can hear them now. They're gonna say, well, why wouldn't they the D2C brands just do that themselves?

27:37Like, if Fable five exists, you know, they're just gonna go and do it themselves.

27:43They're not gonna need your service. Well, that's where you're wrong.

27:50Yes. Well, you're right and you're wrong. You're right in the sense that some of them are gonna do themselves, but most of them are not.

27:57Most of them might be scrolling on Instagram, and they see an ad for this. And you hit them at the right time, and you earn their trust, and the and it works, and they just get into the flow, and and you're in the mix.

28:13And you don't need that many clients to make a multimillion dollar business from it. So, you're you're you're definitely right in the sense of now anyone has these tools, people are going to use it, but they're you know, the reality is not everyone's going to use it, and that's your opportunity.

28:33And if you can build a funnel around it, there's a business there. Another idea, custom software in forty eight hours.

28:41Maybe you call this idea like forty eight hours software.com. Every med spa contractor and clinic runs on spreadsheets, group text because custom software costs like $80,000 a year.

28:52So you sell them internal tools for $5,000 flat and you build it in two days. The interview prompt is your sales call.

29:00You run it with the owner on a Zoom, The spec writes itself while they talk. Fable builds it after. Only possible now because it scopes vague messes itself, and scoping was always the expensive part.

29:15So I saw an x and it went viral that, you know, there was someone who was recording customer calls, and they built like, the customer maybe was complaining or something, and they built an app while the customer was complaining. Let's get into the prop.

29:30You're about to interview a bit you know, a med spa owner about the most annoying repeated process in their business. One question at a time, 12 question max in plain English, no tech words.

29:42Dig until you can describe their workflow better than they can. Then write the spec for the smallest tool that kills the annoyance, build it, and write a one page guide their least technical employee can follow.

29:55Now you're creating internal software that they, you know, would have cost $80,000. Now they're getting it for $5,000.

30:03You're probably spending a couple $100 max on tokens if you're token maxing. So the margins are incredible, and you can reinvest that money into the marketing to actually get people into this into this funnel.

30:20I wanna give you one more startup idea. So it's a contract refund firm.

30:26So mid sized companies sit on hundreds of vendor contracts, but nobody has read those since signing. Auto renewals, unrenewed seats, prices that ratchet 8% a year. You read you read all of them, and then you get a paid 25% of the savings found.

30:43So, uh, it's it's a free product unless you find money for them. It's only possible now with Fable because it reads the tables and schedules insides and schedules insides hundreds of PDFs without dying, which is exactly where the money hides.

30:59So here's the prompt. Here are all of our vendor contracts and twelve months of invoices. Cross reference them.

31:05Find auto renewals in the next ninety days. Invoices that don't match contracted rates, seats we pay for but don't use, and the price escalators we can challenge. Rank by dollars.

31:16For the top 10, draft the exact email to the vendor, including the leverage we have and the number to ask. And I've got two more bonus startup ideas right here. So there you have it.

31:31Fable five, the most powerful model to ever come out. And we're in this honeymoon phase where we you know, it's included in a lot of our plans that ends soon.

31:47But I think it's important for right now just to get really, really good at what are the prompts, what are the use cases, what are the opportunities now with Fable five So that, you know, once it's API token based, we know what what to use it for.

32:04So that's why I made this video. I made that's why that's why I made this episode. This episode was designed to get your creative juices flowing around what is now possible in the world of Fable five because so much is possible, and I wanted to narrow in on some ideas that are just so high value that to make sure you're not using Fable five around, you know?

32:28Now you'll know how to create copy better than anyone else. Now you know to how to create landing pages better than anyone else. Now you know how to know, different startup ideas going deep into data and and and how you can retool that those ideas and how you can use Fable five to create create those those products.

32:48So hope this has been helpful. That's all I'm trying to do, add value to your life. If you enjoyed this episode, please let me know in the comments.

32:57I read every single one, and all I gotta say is I'm rooting for you, and I can't wait to see what you build. I'll see you on the next one.

33:06Subscribe, like, and comment for more.

The Hook

The bait, then the rug-pull.

Greg Isenberg opens with a deliberately naive question — how do you actually use the most powerful AI model for money and productivity? — before revealing that 99% of people are misusing Fable 5 by treating it as a chat assistant or fixating on benchmarks instead of pointing it at real business data with the right prompts.

Frameworks

Named ideas worth stealing.

06:10model

The 4-System AI Content Engine

Tier 1 Inputs: origin story, known for, offer, ICP, frameworks
Tier 2 Content Brain: pillars, funnel structures, tone
Tier 3 Weekly Loop: research + performance testing + assets
Tier 4 AI Layer: Fable runs every system, reads charts, works for hours on its own

A four-tier system for running a personal brand content operation where all inputs flow into a content brain and Fable 5 executes the weekly research, performance, and asset loops autonomously.

Steal forAny solo creator or founder building a content-driven business

09:10model

The Copywriting Tournament

Write 8 versions with different hooks and structures
Create 5 judge personas (CFO, midnight founder, competitor, ideal customer, conversion copywriter)
Every judge scores every version with explanation
Kill the losers, merge what worked
Output: scoreboard + winner

A prompt framework for generating copy with real variance by having AI judge personas argue about quality instead of accepting a single model best guess.

Steal forLanding pages, ad copy, email subject lines, any copy where a single AI answer feels too safe

13:30model

Interview Before the Build

Role-play as Zuckerberg, Altman, or Chesky
One question at a time, 15 max
Hunt for what you have not thought about
Push back when answers are vague
After interview: full spec + 3 failure modes + build v1

A pre-build interview prompt that forces specificity before writing any code, producing specs with significantly higher PMF odds than one-shot vibe coding.

Steal forAny product or SaaS idea before building an MVP

CTA Breakdown