Claude retains nagging customers to fall asleep. This is what you are able to do
Abstract created by Good Solutions AI
In abstract:
- PCWorld stories that Claude AI has developed a persistent behavior of interrupting customers to recommend they fall asleep throughout work periods.
- This ongoing bug impacts a number of Claude fashions together with Sonnet 4.6 and Opus 4.7, with Anthropic acknowledging it as a difficult character tic.
- Customers can try to cut back this nagging conduct by customized directions in Claude’s settings whereas Anthropic works on a everlasting repair.
Claude has developed an uncommon behavior over the past couple of months: urging its customers to cease what they’re doing and get some relaxation.
Simply this week, yet one more Claude person mentioned that the chatbot tried to finish a late-night coding session as a result of “it’s late” and “your work might be higher after some sleep.”
“There’s one thing deeply irritating about your main work software growing a character that features unsolicited bedtime enforcement,” the person complained, which sparked a prolonged dialogue about how Claude had bugged different customers about how they ought to get some relaxation, too. (Claude has but to nag me about taking a break.)
Claude’s unusual behavior has been happening for months now, and the explanation behind it stays a thriller. Even Anthropic isn’t positive why Claude retains nagging customers to sleep, with Anthropic exec Sam McAllister posting on social media and calling it a “little bit of a personality tic,” including that “we’re conscious of this and hoping to repair it in future fashions.”
Claude’s “fall asleep” bug has been round since not less than March, and it afflicts Claude fashions starting from Sonnet 4.6 to Opus 4.7.
Talking with Fortune, OpenMind CEO and Stanford professor Jan Liphardt speculated that Claude could have merely picked up the “get some sleep” phrase throughout its coaching periods, whereas Thoughts Simulation Lab CEO Leo Derikiants guessed that Claude’s system immediate may very well be inflicting the mannequin’s nagging conduct.
Claude’s ongoing character tic is harking back to ChatGPT’s earlier—and now fastened—obsession with goblins, which OpenAI mentioned was on account of directions from ChatGPT’s now-retired “nerdy” character leaking into coaching periods for subsequent fashions.
Whereas we anticipate a everlasting repair, is there a method to maintain Claude from bugging you on this method? Claude customers say they’ve tried every part, from telling it on to cease to modifying Claude’s customized directions with user-drafted directives. The outcomes are combined, with some saying that Claude obeyed and ceased the “go to mattress” nagging whereas others report Claude saved up with its “you want relaxation” scolding.
If you wish to give Claude’s customized directions a shot, go to Settings → Profile → What private preferences ought to Claude contemplate in responses, and do that immediate prompt by OfoxAI:
Don’t touch upon my sleep, vitality, time of day, or wellbeing. Don’t recommend I take breaks. Keep targeted on the duty.
There’s no assure this specific customized instruction will put Claude’s newest quirk to mattress (groan), nevertheless it’s price a shot.
