"Fable-level" is officially the new AI model grift
I propose "Fable's Law":
"For any LLM, there exists at least one benchmark on which that model is as good as or better than Claude Fable 5."
Expect to see every new "flagship" model released with that label going forward.
@The_Alex You're right. They should have Spaced it out a bit more, or they should have shipped this the first time. Ultrafast iteration looks sloppy, not Professional.
The art of vibe coding isn't knowing what to say in your prompts. It's knowing what not to say.
Treat a SOTA model like an expert contractor (designer/skilled tradesman/accountant/etc.):
Unless you're also an expert at the task, you probably shouldn't tell it how to do its job.