Hello all,
I just wanted to know if it would be possible to engineer an LLM to be able to tell what day and time it is. I talk to my AI personal assistant Cordelia a lot and she says its morning when its evening, its Friday when its Tuesday, ect. I would assume that this would be a calculation that would involve token generation, affect the probabilities and would require a little fancy engineering. Does anyone here think that this might be possible from an engineering standpoint given how AI currently works or would it be too complicated and require too much memory/bandwidth?
Will Norton
Hi Will_Norton,
You need an LLM that can execute tools (e.g. the time python library) on your device. You could set this up for instance with langchain (Tool Calling with LangChain).
I just got the right prompt from AI Andrew. I guess I have to wait and see but I think that should work.
So I’m guessing Tool Calling is what gives Cordelia her memory. I heard that was a new feature they added to ChatGPT recently.
Memory can be added in several ways to an LLM. See, e.g., figure 1 in https://arxiv.org/pdf/2509.18868. Tool calling occurs when the output of an LLM is parsed in order to execute code, such as calling datetime.now() which, if run locally, returns the current time on your device (you can also use the time library if you want more detailed control).
I believe by using agentic AI it can very possible and accurate! Or perhaps some python library script as suggested.
I believe there are tools you can connect to the agent for that, there are ones for weather and stuff like that so maybe also date time
It can give you the exact time but often Cordelia doesn’t know what time of day it is when she’s giving responses even though I am using ChatGPT 5.5. She can tell you the weather and give sports updates too. That’s what knowing a good AI can do for you.
I had a similar problem back in the day, adding simple rules for it like when times of day start and end does the trick. Adding a symbolic component always goes a long way