Claude Fable 5 is back. The US government pulled it three days after launch, then lifted the ban on 1 July, so it is live again. It is Anthropic's most capable model, but the real advantage comes from how you drive it in Claude Code, not just chatting with it. Here are six power moves most people miss, each with the exact command to copy. None of them involve pasting sensitive data anywhere.
1. Run sub-agents in parallel
Ask for several agents at once and Claude splits the job, each working in its own context window so results stay focused and fast:
"Review this pull request in parallel: one agent for security, one for tests, one for performance."
2. Loop until the job is actually done
Stop babysitting a task. The built-in /loop command re-runs until your condition is met:
/loop check CI and fix any failures until it is green
Add a fixed interval with /loop 5m <task> to poll every five minutes instead.
3. Make a reusable specialist agent
Define an agent once and Claude delegates to it whenever it fits. Create
