I'm trying out Claude Code for the first time and it's been an exciting experience! It's a solid coding tool, especially with the Max subscription providing reliable costs. However, I have a question: Why doesn't it include the current system timestamp in every message? This feature could make things so much simpler without using up many tokens. Right now, I have to tell it to run a system date command before writing dev logs, which sometimes leads to it creating random timestamps instead of using the actual time. I'd love to see this feature added without caching the timestamp. Thanks for considering this suggestion; I'm really impressed with Claude Code and can't wait to see how it develops!
2 Answers
Timestamps are a tricky deal because LLMs work within static context windows, which don’t change with time. It's different from how OpenAI does it; they dynamically inject the current time into their system prompts, so while the model itself might be static, the context includes the real-time data when needed.
You might want to consider wrapping your dev log writing in a script that includes the timestamp. I get why it might seem like a simple feature, but adding timestamps could complicate prompt caching for edge cases. Claude Code aims to support a wide range of uses, following a broad approach similar to the UNIX philosophy. It could be useful to put in a feature request via their public GitHub. That way, they know there's interest!
That's interesting to hear! I appreciate the insight into how other models handle it. It would be super helpful if Claude could adopt something similar.