Claude Fable 5 Finds Rules Against Slop

The useful takeaway is not that Fable “writes human,” but that detection systems built around visible patterns are easy to game.

• –The Slop Index measures conciseness, templating, rhythm, stylistic tells, and human preference across 112 scenarios.
• –Fable’s ability to generate detector-beating rules exposes the weakness of static checklists and surface-level classifiers.
• –Better writing tools should encourage specificity, original thinking, and meaningful revision—not merely remove phrases associated with AI output.
• –AI detection remains a poor proxy for authorship because human writing can trigger the same patterns, while AI-generated text can be deliberately optimized around them.
• –The stronger product direction is transparent, quality-focused assistance that preserves the writer’s voice instead of hiding AI involvement.