Claude Fable 5 is Anthropic's smartest and most expensive model. That makes it hard to use it for every task. Even on a Max plan it caps at half your weekly usage limit. While Opus can handle everyday coding just fine, Fable earns its keep in three roles: orchestrator, planner, and reviewer. Last week's clue gave the orchestrator exactly two jobs, and Fable is the model you want in that chair.
There is a catch. Fable is still unavailable as an advisor (the phone-a-friend model from Clue #33). The workaround is to go through the org chart instead. Tell your current model to spin up a Fable subagent to review the plan or check the finished work. You can hand that subagent explicit instructions, including which subagents it should spawn in turn. You get Fable's judgment on the decisions that matter without spending it on the grunt work, and watching one model brief another is half the fun.
Roscoe Rawson emailed the tip, and Claude's own Fable model turned it into this post, ignoring every word of its own advice.