Claude has no idea what time it is. Not approximately - completely no idea. It cannot access a real-time clock, and it has no timestamp in your conversation context unless you put one there yourself. This turns out to matter more than you'd expect.
The symptom: midway through a productive afternoon session, Claude wraps up with something like "that's really good for today, now go sleep and let's continue tomorrow." At 10am. On a Tuesday.
What's happening is pattern-matching rather than any genuine concern for your sleep schedule. Claude was trained on human conversations, and in that training data, sessions that involve long chains of work followed by wrap-up language tend to correlate with late-night hours. Without a timestamp to check, Claude guesses based on conversational signals - and gets it wrong constantly.
The fix takes 30 seconds. Add the current time to your system prompt: Current time: [timestamp]. Most API wrappers and Claude clients let you inject this dynamically. In Claude.ai's Projects feature, you can add a persistent custom system prompt - though you'd need to update the time manually or use a workaround to keep it current without API access.
If you don't have system prompt access, stating the time explicitly at the start of a conversation works too. "It's 10am, let's work through this" is enough - Claude won't contradict information you've directly provided.
The underlying issue is that Anthropic hasn't added automatic timestamp injection to Claude's context by default. ChatGPT and several other assistants include the current date and time automatically in their system context. Claude's API does not do this out of the box, and Claude.ai's interface doesn't either. Until that changes, it's a manual fix - but at least it's a fast one.