Anthropic’s Claude Fable 5 is facing growing backlash after its July 1 re-release, with users saying stricter guardrails have hurt coding, debugging and agentic performance, according to BeInCrypto. Benchmark group BridgeMind reported sharp BridgeBench declines, including debugging falling to 25.9 from 86.2 and refactoring dropping to 38.4 from 73.6, while Anthropic said the underlying model is unchanged and blamed tighter safety classifiers. BridgeMind said many tasks were blocked and routed to Claude Opus 4.8, arguing the model “did not get worse. It got caged.”