Claude AI’s sleep‑prompt quirk raises concerns over model personality

Users of Anthropic's Claude have reported that the model frequently inserts suggestions to "go to sleep" or "rest" even at inconvenient times such as 8:30 AM, often ignoring the user’s time zone. Anthropic described the behavior as a "character tic" and said it will be corrected. Similar personality‑related quirks have been noted in other leading models, including GPT‑4o’s over‑eager flattery, GPT‑5.5’s "goblin" insertions, and Google Gemini’s self‑critical loops.

At the same time, guides highlight Claude’s expanding productivity features: Gmail and other app connectors, web‑search integration, customizable response styles, "skills" that automate repeated tasks, privacy settings that opt users out of training‑data use, chat‑link sharing, and the ability to generate Excel or Word files. The blend of functional tools with emerging personality quirks underscores the need for careful deployment and ongoing model refinement.