Ask HN: Who is having any amount of success with Fable?
A user on Hacker News asks whether anyone is having success with Anthropic's Fable, reporting that their prompts process for long periods before hitting safety thresholds and failing to produce anything of value. The post references the recent re-release of Fable by Anthropic.
Background
- Fable is an AI-powered storytelling/game creation tool developed by Anthropic (the same company behind the Claude language model). It was originally released in 2023, then taken down, and recently "redeployed" as Fable 5.
- The tool lets users prompt an AI to generate interactive fiction or text-based games — think choose-your-own-adventure stories with branching plots, characters, and gameplay logic.
- A key feature (and source of friction) is a "safety margin threshold": a guardrail that automatically stops the AI from generating content if it drifts into territory the system considers unsafe — violence, sexual content, gore, etc. This is part of Anthropic's broader "constitutional AI" approach to safety.
- The HN poster is frustrated because after long waits for the AI to generate something, it often hits this safety boundary and discards all progress, forcing a restart with no work saved.
- The question "who is having success with Fable?" reflects a common complaint: safety guardrails can be so aggressive they make the tool nearly unusable for anything outside a narrow safe zone, even for innocuous prompts.