I used Fable 5 until the weekly limit and realized: Most people can finish everything with Sonnet 5
Fable 5 was, honestly, amazing.
Coding.
Video editing.
Anime production.
Writing.
I used it so much that I hit the weekly limit in just one day.
To put it simply, it's the best. It was so good it almost blew my mind.
The code is persistent.
The writing is deep.
It can create video structures.
It can even output anime cut-lists.
It picks up on my vague intentions and takes them quite far.
You'll know it when you touch it.
This thing is powerful.
But after using it to the limit, I actually thought the opposite.
Most people don't need Fable 5 yet.
This isn't to say Fable 5 is weak. Quite the contrary.
It's too strong.
Most people aren't doing work that requires Fable. Before that, they aren't even using Sonnet 5 properly.
Many people ask Claude like this:
"Write an article."
"Make a presentation."
"Think of a video structure."
"Fix this code."
"Create an LP."
Even with this, something "decent" comes out.
But it usually stops at just being "decent."
Why?
Because they are using Sonnet 5 as a regular chat.
Fable 5 is strong not just because it's smart. It thinks longer. It breaks down tasks. It doubts itself midway. It proposes alternatives. It destroys its own answers. Finally, it integrates everything.
This "persistence" is its strength.
So, we should just force Sonnet 5 to follow the same process.
Sonnet 5 used normally is just a talented AI. But Sonnet 5 in "Extreme Activation" mode, with a structured process, gets remarkably close to Fable.
Why Ordinary Prompts Are Weak
I previously shared a prompt like:
"Divide the work among a Strategist, Editor-in-Chief, Practitioner, Inspector, and Distribution Lead."
The direction wasn't bad. But it's still weak.
Because it only divides roles.
Even if you divide roles,
If there are no evaluation criteria, it stays shallow.
If there is no rebuttal processing, it feels fake.
If there are no multiple drafts, it gets pulled by the first idea.
If there is no self-grading, it won't improve.
If there are no finishing conditions, it's unusable.
If you want to approach Fable-level, you don't need "5 subordinates."
You need to run those 5 subordinates through 3 cycles.
Cycle 1: Create the direction.
Cycle 2: Destroy the weaknesses.
Cycle 3: Finish the product.
Only after doing this does the output truly change.
The Framework to Bring Sonnet 5 to Fable-Level
You use these 5 roles:
Strategist.
Designer.
Creator.
Destroyer.
Finisher.
Here is where it differs from before.
You don't just call them in order. You give each of them scoring criteria and redo conditions.
Command Claude like this:
"From now on, do not act as a normal chat AI, but as a high-load production team.
The goal is not to 'answer quickly.'
The goal is to 'create a deliverable that can be used as-is.'
Execute the following 5 roles in order:
- Strategist
- Designer
- Creator
- Destroyer
- Finisher
Furthermore, do not stop at one pass; you must complete 3 cycles.
Cycle 1: Establish direction
Cycle 2: Destroy weaknesses
Cycle 3: Finalize the form
In each phase, you must output:
- Judgment
- Reason
- Missing information
- Failure risks
- Improvement plans
- Next steps
Note:
Vague generalities are prohibited.
AI-like platitudes are prohibited.
Half-baked prompt collections are prohibited.
Make it a deliverable that can actually be used."
With this, Sonnet 5 finally enters "Thinking Mode."
But it's still not enough.
The real strength starts here.
First, Create "Victory Conditions" Instead of a "Winning Strategy"
Most people have Claude start creating immediately.
This is weak.
The first thing to create is not the deliverable. It's the Victory Conditions.
For an article, it looks like this:
"First, do not write the body text.
Define the conditions that would constitute a 'win' for this theme.
Theme:
I used Fable 5 to the limit and realized: Most people can finish with Sonnet 5.
Reader:
People using Claude or AI for work. People curious about Fable 5 but unable to judge if they need it.
Purpose:
While acknowledging Fable 5's greatness, convince readers that improving their use of Sonnet 5 comes first.
Output:
- Reasons readers will click
- Reasons readers will save
- Points readers will argue against
- Responses to those arguments
- The strongest claim
- Claims that must NEVER be made
- Victory conditions for the article
- Scoring criteria out of 100 points"
Have it output this first.
What's important here is having it output the "Claims that must NEVER be made."
In this case, things like:
"Fable 5 is weak."
"Fable 5 is not worth using."
"Sonnet 5 and Fable 5 are exactly the same."
"Anyone can do everything with just Sonnet 5."
These sound fake.
The correct approach is:
Fable 5 is the best. But most people haven't fully utilized Sonnet 5 before moving to Fable.
This claim is strong because it's about the order of operations, not a denial of quality.
To Approach Fable-Level, "Create 3 Drafts, Then Kill Them"
People with weak Claude output usually use the first draft.
The first draft is weak because Claude outputs the safest answer first.
So, ask it this:
"Next, provide 3 directions for the same theme.
Plan A: The easiest to understand
Plan B: The most edgy/sharp
Plan C: The most likely to be saved
For each, output:
- Title
- Intro
- Main claim
- Reader's takeaway
- Strengths
- Weaknesses
- Viral potential
- Risk of backlash
Then, compare the 3 plans and choose the one that should ultimately be adopted.
However, do not escape into an easy 'average' plan. Choose the plan with the highest potential for growth."
Doing this changes the output instantly.
Normal Claude answers immediately. Strong Claude creates options and then chooses.
If you want to approach Fable-like output, you need to make Sonnet 5 compare.
The Most Important Role is the "Destroyer"
This is the main point.
To get Fable-level results from Sonnet 5, the most important role isn't the Creator. It's the Destroyer.
Most people stop after the AI creates something.
But in reality, AI becomes strong after it has created.
Create. Doubt. Destroy. Fix.
The output grows through this back-and-forth.
So, always ask this:
"You are now the Destroyer.
Do not praise the current deliverable. No need to look for good points.
Destroy it ruthlessly based on these criteria:
- Does the intro really make people want to keep reading?
- Is the claim strong, or does it sound fake?
- Does it anticipate reader rebuttals?
- Are the concrete examples weak?
- Where does it sound like an AI?
- Where is it boring?
- Where will readers drop off?
- Which words feel 'light' or shallow?
- what should be cut?
- What should be added?
Output format:
- Fatal flaws
- Weaknesses
- Parts to cut
- Parts to add
- Rewrite suggestions
- Re-scoring out of 100"
You must include this.
Just saying "Review this" is weak. Claude is polite, so it will praise you. Praise doesn't make the work stronger. You need destruction, not a review.
Finishing Means Doing It Until It's "Ready to Publish"
Most AI usage is weak at the end.
The body is done, but the title is weak. The thumbnail text is weak. The quotes are weak. There are no video references. There is no post text.
This won't grow.
In an X article, the quote-repost text is effectively the thumbnail. If it's a short quote, it needs to make people want to click; if it's a video quote, it needs to make people want to watch. This should be crafted as much as the body.
So finally, ask this:
"You are the Finisher.
Finish this deliverable to a state where it can actually be published.
Output:
- 10 Final title ideas
- 10 Thumbnail copy ideas
- 10 Self-quote ideas
- 5 Video quote ideas
- 5 Demonstration scenes to show via screen recording
- 5 Versions of the intro made even stronger
- 10 'Save-worthy' sentences
- 5 Discussion points to encourage comments
- Pre-posting checklist
- Final version
Conditions:
- Do not duplicate title and thumbnail text
- Keep thumbnail text short
- Prohibit AI-like expressions
- Don't make quotes half-baked long sentences
- Make video quotes enticing
- Ensure the body text has practical utility worth saving"
Only after this is it complete as an article.
The Archive Edition: Sonnet 5 Extreme Activation Prompt
This is the main event. Please keep this prompt from the article.
"You are not a normal chat AI. Act as a high-load production team to reproduce 'Fable-level persistence' in Sonnet 5.
The goal is not to answer quickly. The goal is to create deliverables that can be used as-is.
Act as the following 5 subordinates in order:
- Strategist: Analyzes market breadth, specificity, reader desires, counterarguments, and the path to victory.
- Designer: Breaks down the process to completion, organizing necessary info, work order, and failure risks.
- Creator: Makes the actual deliverable. Doesn't settle on one draft; always outputs 3 for comparison.
- Destroyer: Destroys the creation without praise. Points out boring, AI-like, fake, or weak parts.
- Finisher: Prepares the final deliverable for publishing/delivery/posting.
You must complete 3 cycles.
Cycle 1: Create direction
Cycle 2: Destroy weaknesses
Cycle 3: Finalize form
In each cycle, output:
- Judgment, Reason, Missing info, Failure risk, Improvement plan, Next steps, Self-score/100.
Final output must include:
- Final deliverable, Rejected plans and reasons, How weaknesses were crushed, Pre-publish checklist, Next instructions for Claude.
Theme: [ ]
Purpose: [ ]
Reader: [ ]
Final Deliverable: [ ]
Constraints: [ ]
Start with 'Victory Conditions' and 'Scoring Criteria' before creating the body."
This is the usage that separates you from those who use Sonnet 5 normally.
It's not "Write an article." It's "Force the thinking process until completion."
It's not "Make a presentation." It's "Run through victory conditions, structure, destruction, and finishing."
It's not "Write code." It's "Do requirements, design, implementation, testing, README, and bug crushing."
This is Sonnet 5 Extreme Activation.
How to Use by Category
Writing
Theme: I used Fable 5 to the limit... (etc.)
Final Deliverable: X article, title, thumbnail, self-quote, video quote.
Conditions: State "Fable was great" at the start, then pivot to "Most people don't need it." Don't deny Fable. Provide the Sonnet 5 activation method as a saveable resource.
Coding
Purpose: Implement [ ] function.
Final Deliverable: Implementation policy, file structure, code, test points, bug fix ideas, README.
Conditions: Don't write code immediately. Identify specification ambiguities first. Compare 3 implementation policies.
Video Production
Purpose: Make a video on [ ] theme.
Final Deliverable: Planning, first 3 seconds, script, cut-list, captions, BGM direction, post text.
Conditions: Hook them in 3 seconds. Create retention points every 15 seconds. Cut boring explanations.
Anime Production
Purpose: Create a short anime with [ ] worldview.
Final Deliverable: Worldview, character settings, scene structure, shot list, camera instructions, movement, dialogue, AI prompts.
Conditions: Don't make images immediately. Decide the worldview and emotional axis first. Clarify the role of each cut.
Conclusion
Fable 5 was the best.
I used it for coding, video editing, anime, and writing until I hit the limit.
Honestly, it was mind-blowingly good.
But I'll still say it.
It's not necessary for most people yet.
Because most people haven't reached the point where they need Fable.
Before using Fable 5, have Sonnet 5 create victory conditions. Have it break down tasks. Have it output 3 drafts. Have it destroy its own work. Run 3 cycles. Finish it to a publishable state.
If you do this, most work can be finished with Sonnet 5.
Normal Sonnet 5 is just a talented AI.
But Extreme Activated Sonnet 5 is different.
You can open Fable after you've done that.
Seal Fable. Complete it with Sonnet 5.
