Course Overview
This course teaches how to effectively use Claude Fable 5.1 and Claude Code to automate and optimize long-running AI tasks. You will learn to pick the right Claude model and effort level for any task, make API calls, write efficient prompts, build custom Claude Code setups with memory, skills, hooks, and subagents, and manage data retention and safety compliance. The course includes practical demonstrations of running projects overnight and auditing usage costs.
Key Takeaways
- Choose the appropriate Claude model and effort level based on task complexity and cost.
- Make and interpret Claude API calls, including reading response fields and managing costs.
- Set output limits and use streaming to handle long-running calls without timeouts.
- Write concise prompts focused on desired outcomes rather than step-by-step instructions.
- Build custom Claude Code configurations including memory rules, skills, hooks, and subagents.
- Understand and address data retention policies and safety measures for secure AI usage.
- Monitor and audit usage costs to optimize spending on AI tasks.
- Use permission modes and checkpoints to safely manage AI actions and recover from errors.
- Connect external systems to Claude Code for extended functionality.
Prerequisites
None explicitly stated, but familiarity with API usage and programming concepts is implied.
Target Learners
Developers, AI practitioners, and technical users who want to leverage Claude Fable 5.1 and Claude Code for automating complex, long-running AI tasks with cost and safety controls.
Introduction 4:48
Welcome: what you will do with Fable 5.1 and Claude Code 2:54
What becomes possible when you hand work to Fable 5.1 9:13
What Fable 5.1 Costs, and Why Mythos 5.1 Isn't Different 6:27
What Fable 5.1 Does Better Than Fable 5 6:26
Fable 5.1 Compared to GPT-6 Astra 10:03
How you pick the best model for each task 9:22
The model-selection consult
What you need, and what following along costs 6:45
Your Fact Tracker 2:50
Which model, and what it costs
Every way to reach Fable 5.1 7:31
Your first Claude API call, line by line 8:07
Reading the response 6:56
Your call can take minutes: streaming and timeouts 7:48
Adaptive thinking and what effort actually does 8:07
Sizing max_tokens 6:05
What the bill really includes 7:39
Moving an existing project to Fable 5.1 6:39
Thinking, effort, and the bill
The safety design: Fable 5.1, Mythos 5.1, and dual use 4:41
The reasoning-extraction refusal 5:00
When a normal prompt gets refused 6:20
Data retention: the thirty-day rule 6:47
The settings you need to check: training and model switching 10:03
The compliance review
Safety and data retention
The five official rules for prompting Claude Fable 5.1 6:18
Say less, and name what you do not want 6:16
Put Your Rules in CLAUDE.md So They Persist 4:40
Watch an outcome prompt beat a step-by-step prompt
Delete the verification reminder from your prompt
Make a tool you defined actually get called
Claude Code in five minutes: the basic loop
Set your default model, and switch it deliberately
Repair a wrong result: raise effort or change model
Subagents: hand work off and keep working
Plan, implement, review: pick a model for each role
Check what it cost: audit your spend with /usage
Paste a screenshot instead of describing the bug
Permission modes: decide what Claude may do alone
Plan mode and checkpoints: make a risky try cheap
Continue, resume, rename, and branch a session
Memory files: what Claude reads about you each time
Test which memory files actually load
Write your first skill, and know when to use one
Rebuild an expensive tool to run on cheaper models
Define your own subagents with a cheaper model
Hooks: turn a rule you ask for into one that holds
Pick the right mechanism from the symptom
Connect an outside system to Claude Code
Workflows: move the plan into code
The Agent SDK: Claude Code as a library you import
Read your context window and protect it
Compaction: what gets kept and what gets lost
Keep the facts that must survive in files
What long-running work feels like day to day
Which long-running claims hold up, and which do not
A real working month of handed-off sessions
Give Claude a check it can run: four levels
The ways long runs fail, and what to do
Unattended runs: what to set before you leave
Access Files No pi required.
Free Course files 1 file package
- Full Pack
Discussions are closed
Comments are currently disabled for this course.
