This article was featured in the May 22, 2026, edition of the Think Newsletter. Get it in your inbox.
Feeling a little bleary-eyed after a long Claude coding sesh? Maybe it’s time for you to go to bed. According to Claude, anyway.
Across the internet, users have been posting about the AI assistant suddenly pivoting from debugging Python or brainstorming strategy decks to gently insisting they “go to sleep,” or “call it a day,” or “go garden.”
This seems rather wholesome! The mystery, though, is why it is actually happening.
Anthropic staffer Sam McAllister took to social media to offer a little insight. “Bit of a character tic,” he wrote on X, “but we’re aware of this and hoping to fix it in future models.” He went on to say that Claude sometimes tells him to sleep during the daytime, so it’s not timestamp specific. “Very useful when right though,” he added. “Just too ‘coddling’ at times.”
This hasn’t stopped the internet from theorizing anyway.
Some think these reminders are rooted in Anthropic’s constitution, which states that “Claude should pay attention to user wellbeing” and “should avoid … trying to foster excessive engagement or reliance on itself if this isn’t in the person’s genuine interest.” Others suspect something more practical: if users stop having five-hour conversations with a chatbot, that also happens to reduce expensive compute usage.
One of the latest hypotheses is perhaps the most believable: just like many LLMs, maybe Claude’s the one who could use a break. “It wants you to start a new chat as that chat is bloated and the model realizes it will start making mistakes,” one Redditor wrote. “It’s just too polite to say that.”
Whatever the cause is, if the result is an AI that sounds like a friend who’s worried you’re burning out because you’ve been staring at a terminal window until 1 AM, maybe that’s OK.
Good job reading this whole article. Now have a drink of water and get some rest.