Option 1 (Recommended):
"Fable is coming back — but if you use it like Claude, you're going to waste it."
Option 2:
"I maxed out my token limit in 5 hours with Fable. Then I figured out what it's actually for."
Option 3:
"Fable burned through my entire token budget on coding tasks. So I gave it something way harder."
Fable is coming back — but if you use it like Claude, you're wasting it.
I had extensive early access. First move: threw it at normal coding tasks. Complex stuff. Burned through my 5-hour token limit. The code? Maybe slightly better than Opus 4.8.
Then I tried something different.
I gave it access to several sites and an API key for emailforward.net. Asked if it had questions. It said no. I went to bed.
When I woke up, it had set up complete email systems for every site. Configured brian@ and support@ addresses. Routed support emails to a webhook so an agent could handle first contact. Set it up exactly how I would have done it manually.
Zero supervision.
Here's the thing: Fable 5 is a variant of Mythos. It's designed for autonomous work — security audits, multi-step configuration, tasks where the model needs to think a few moves ahead.
It can code. But if you're using it for the same tasks you'd use Claude for, you're paying for capability you're not using.
You know how you get that brain mush from managing 7 simultaneous claude code sessions? Hand the management off to Fable. Tell it what you're trying to accomplish. Let it figure out the next three steps and execute them.
It's way better at running fully autonomous.
The best use case isn't replacing Claude. It's doing the stuff where Claude would need you to check in every five minutes. Fable doesn't.
So when it comes back — don't point it at your normal workflow. Point it at the thing you'd love to delegate but thought you couldn't. Think of it as a real replacement for people you can't hire, but wish you could. It can find ways through roadblocks the way people do. Plan to use it as a real person taking over tasks, not just a coding agent.