5318008

🚨Claude is telling users to go to sleep mid-session, and nobody, including Anthropic, seems to fully understand why it keeps doing it.

— Harshit Jain (@jain_harshit) May 24, 2026

I've remarked previously on my own tools' penchant for telling me to stop working, so this isn't surprising in the least.  And the "nobody understands why" stuff is really silly because it's clear that LLMs have been trained on narrative arcs that lead them to "sense" the natural end of an interaction, and thus signal to you that you ought to go away.

There is no deeper mystery here.  Folks who are puzzled by that likely also think this is some kind of dark magic:

The fucking things are just behaving as they've been designed.

Selah.