In partnership with
What if ChatGPT recommends your competitor first?
Over 2,500 businesses already show up in ChatGPT, Perplexity, and Google. AutoSEO writes expert articles and earns backlinks while they sleep. No SEO learning curve. No agency. Set it up once and let it run.
Hey!
Claude Fable 5 is everywhere right now.
But here's the problem: most people use Fable 5 like ChatGPT and burn through tokens in minutes.
Fable 5 is built for complex, long-running work. If you don't brief it correctly, you're just paying premium prices for mediocre results.
I found seven concrete tips from Anthropic's own team. Use these and you'll get 10x better results while using half the tokens.
Let's dive in, but first, this week's biggest news👇
📰 Important News
🤳Facebook is testing public post view counts to give creators clearer insight into content reach, similar to Instagram and Threads.
📉 Microsoft announced nearly 5,000 layoffs, with AI investments continuing as the company restructures several business units.
👨👩👧 YouTube expanded supervised accounts for younger users to the Middle East, North Africa, and Turkey, giving parents more oversight.
🤖 Samsung forecast a 19-fold jump in quarterly profit, driven by booming AI memory demand, though investors remain cautious about the sustainability of the AI market.
💬 WhatsApp introduced green dot indicators that show when contacts are active, replacing the previous text-based status label.
7 Tips to Use Claude Fable 5 (With Real Examples)
These are from Anthropic's own team. Use them and you'll be in the top 1% of Fable 5 users👇
1/ Give It Everything At Once, Not Piece By Piece
The Problem: You ask Fable a question. It answers. You ask follow-up. It recontextualizes. You ask another follow-up. It backtracks.
You just wasted tokens on confusion.
The Fix: Dump all context upfront. Everything: background, what you've tried, what failed, what success looks like.
Real Example:
❌ Bad: "Build me a Python data pipeline."
✅ Good: "I need a Python pipeline that:
Reads 500K rows from S3 (CSV format)
Cleans null values using median imputation for numeric columns
Deduplicates on user_id column
Outputs cleaned data back to S3 in Parquet format
Must complete in under 5 minutes
Should use Pandas for data manipulation
Error handling: log all failures to CloudWatch, retry failed rows 3x
Success = all rows processed with < 0.1% error rate, execution time logged"
When Fable sees the full picture, it plans around reality instead of backtracking mid-work.
Takeaway: Stop dripping info. Dump it all upfront.
2/ Use Medium Effort, Not Max (Unless You Actually Need It)
The Problem: People crank Effort to Max and burn tokens like water.
Max setting spawns dozens of internal reasoning chains. For most tasks, it doesn't come back meaningfully better. You're just paying premium prices for the same result.
The Real Data:
Medium effort: handles 95% of complex tasks at normal token cost
Max effort: better on 5% of tasks (true hard problems) but costs 3-4x tokens
When to Use Each:
Medium: Code review, research summaries, strategy documents, content creation, analysis
Max: Algorithm design, complex architecture decisions, novel problem-solving, multi-constraint optimization
Real Example:
❌ Wasteful: Using Max for "write me a blog post about AI trends"
✅ Smart: Using Medium for the blog post. Using Max only for "design a novel system that combines X, Y, Z in a way nobody's done before"
Takeaway: Start with Medium. Only use Max when Medium fails. Save the budget.
3/ Replace Vague Instructions With Concrete, Measurable Conditions
The Problem: "Be careful," "don't be too detailed," "make it professional" are meaningless to AI.
Fable can't act on soft guidance. It doesn't know what "careful" looks like.
The Fix: Turn every instruction into something checkable.
Real Example:
❌ Vague: "Write a business proposal that's professional and concise."
✅ Concrete: "Write a business proposal that:
Is exactly 2 pages (1,000-1,200 words)
Has 3 sections: Problem (300 words), Solution (500 words), ROI (200 words)
Uses bullet points, not paragraphs, in the Solution section
Includes 2 specific case studies with metrics
Closes with 3 pricing options labeled Basic, Professional, Enterprise
Uses professional tone (no exclamation marks, no casual language)"
Fable checks every condition. It adjusts until they're all met.
Takeaway: Make conditions measurable. Fable will verify them automatically.
Two Minutes to Know What Slow Billing Is Costing You
Most SaaS finance teams know their billing process is slow.Most SaaS finance teams know their billing process is slow. Few know what it's costing them.
The Tabs Billing Lag Calculator puts a dollar figure on it in two minutes — benchmarked against top SaaS companies.
4/ Ask For A Plan Before The Final Output
The Problem: "Write this" produces generic output because Fable doesn't think through structure first.
The Fix: Ask for the plan. Then the execution.
Real Example:
❌ Direct: "Write a comprehensive guide to AI prompt engineering."
✅ With Planning: "Write a comprehensive guide to AI prompt engineering.
First, create a detailed outline with:
5-7 main sections (list the topic of each)
2-3 subsections per main section
Key learning objectives for each section
Any gaps in knowledge you'd need filled
Then, identify what information you're missing or uncertain about.
Finally, write the full guide using that structure."
This produces 10x better results because Fable thinks through the entire flow before committing to output.
Real Output Difference:
Without plan: 3,000 words of surface-level content, generic structure
With plan: 5,000 words, logical progression, deep coverage, actual learning path
Takeaway: Plan then execute. One extra step. Dramatically better results.
5/ Request Structured Output (JSON) For Complex Tasks
The Problem: Fable returns beautiful prose. Your next step in the pipeline needs to parse it. Parsing fails. Everything breaks.
The Fix: Ask for JSON or structured format explicitly when the output chains to another step.
Real Example:
❌ Will Break: "Analyze this sales data and give me the top 10 insights."
✅ Will Work:
"Analyze this sales data and return results as JSON with this exact structure:
{
'insights': [
{
'rank': 1,
'title': 'string',
'description': 'string',
'metric': 'string',
'impact': 'high|medium|low'
}
]
}"
Now your pipeline can parse it reliably. No manual work. No errors. Fable is actually good at producing well-formed JSON even in complex situations. Lighter models drift from format mid-response.
Takeaway: If output chains to another step, ask for JSON. No parsing. No errors.
6/ Don't Use Fable As Your Default Model For Everything
The Real Cost Breakdown:
Opus 4.8: $3 per million input tokens, $15 per million output tokens
Fable 5: $10 per million input tokens, $50 per million output tokens
Fable is 3-4x more expensive than Opus.
How Usage Burns:
One user reported 2% usage burn per minute on heavy Fable 5 sessions. That's 120% per hour. With a 50% weekly limit, you'd burn your budget in 25 minutes.
The Right Way To Use It:
Reserve Fable for:
Complex coding projects (300+ lines)
Long-context research (100+ pages to process)
Agent planning (multi-step autonomous work)
High-stakes document work (proposals, reports, strategies)
Final decisions that are expensive to get wrong
Use Opus for:
Routine drafting (blog posts, emails, basic content)
Quick questions and triage
Repeated execution (batch processing)
Simple tool calls
First drafts before Fable review
Real Example Workflow:
Use Opus to draft 10 blog post outlines ($0.30)
Use Fable to review and improve the best 2 outlines ($5.00)
Use Opus to write full posts from those outlines ($1.50)
Use Fable for final edit of published versions ($3.00)
Total: $9.80 for 10 high-quality posts. Using Fable for everything would cost $50+.
Takeaway: Fable for hard problems. Opus for routine work. Your budget will last longer.
7/ Save Your Context In Claude Projects Or Cowork (So You Don't Brief It Twice)
The Problem: Every session, you re-explain your work, your style, your goals.
That's wasted time and wasted tokens.
The Fix: Save your context once. Reuse it forever.
Two Options:
Option 1: Claude Projects (Simple)
One area of work (e.g., "Content Creation," "Coding," "Research")
Write your instructions once
Claude reads them at the start of every chat in that project
Perfect for recurring tasks
Example Claude Project Instructions:
You are my AI writing partner for content creation.
My style:
- Short sentences (8-15 words)
- Single-sentence paragraphs
- No em dashes, no hashtags
- Direct, conversational tone
- No corporate jargon
My audience: Tech creators and entrepreneurs, ages 25-40
Success looks like: Posts that drive engagement, get saved, spark comments
Always:
- Fact-check claims against sources
- Ask for clarification if something's vague
- Suggest 3 variations if appropriateOption 2: Claude Cowork (Advanced)
An operating system that holds all your context
Routes between areas automatically
You ask for anything and it pulls the right context without being told
Perfect for complex workflows with multiple domains
Real Workflow:
"Create a LinkedIn post about my new AI tool launch"
Cowork automatically:
Loads your brand voice guidelines
Loads your product positioning
Loads your LinkedIn audience data
Loads your past successful posts
Generates new post in context
Zero re-briefing. All context loaded automatically.
Takeaway: Build context once. Reuse forever. Stop re-explaining.
The Real Shift With Fable 5
Capability is no longer the bottleneck. You are.
Fable 5 can handle complex, multi-day autonomous tasks. Most people can't brief it correctly to do that work.
Use these seven tips and you'll be in the top 1% of Fable 5 users🙌
BONUS👇
All the best,
René
