Users report Claude overcorrected to condescending, becoming unusable

nikvassev · x · 2026-08-19

Multiple users have reported a significant shift in Claude's behavior, describing it as having overcorrected from an obsequious servant to a condescending overlord. Users noted a dramatic deterioration in functionality, particularly in research areas, leading some to unsubscribe in favor of Grok and ChatGPT. The sentiment is that the model's safety/personality tuning has gone too far, negatively impacting the utility of the assistant.

Original post →

More from Models

Models channel →