A few days ago, ChatGPT wished me good night.
There was just one problem: it was the middle of the day.
Of course, an AI can tell you the current time. Ask it, and it can look it up or use the information available to it.
But that made me think about a slightly different problem:
Knowing the time isn't the same as having a sense of time.
When we talk to another person, time is naturally part of the conversation.
We know whether it's morning or evening. We know whether we've been talking for five minutes or two hours. We understand that someone replied immediately — or came back after three days.
That temporal context influences the conversation without either person having to explicitly mention it.
With an AI conversation, that context can be missing or inconsistent.
So I decided to experiment with giving AI something very simple:
A watch.
What AI-Watch does
I built AI-Watch, a browser extension that automatically provides temporal context to your AI conversations.
It can provide:
Local date and time
Timezone
Current session duration
Time elapsed since the previous exchange
The idea isn't to create another clock.
The idea is to make time part of the conversation context.
And AI-Watch isn't tied to ChatGPT. I designed it to work with major AI assistants, because the underlying idea isn't specific to one model.
Privacy was non-negotiable
I didn't want something this simple to require another account, another cloud service or another piece of tracking software.
So AI-Watch runs locally.
No account.
No tracking.
No telemetry.
And it's completely free.
Then I added something completely unnecessary
While building it, I started thinking about the object itself.
If you're giving an AI a watch, shouldn't the AI be allowed to have an opinion about its gift?
So I added a button:
“What do you think of your gift?”
Press it and the AI can react to the watch you just gave it.
The answer isn't prerecorded. The AI generates its own response, so different assistants — and even different conversations — can react differently.
Is this feature essential?
Absolutely not.
Is it my favourite part?
Probably.
A small experiment around a bigger question
AI-Watch started with a funny interaction, but it led me to a question I find much more interesting:
How much contextual information do AI conversations still lack simply because humans take that information for granted?
Time is one example.
Humans don't constantly announce the date, the hour or how long it's been since the previous sentence. We simply understand those things as part of the situation.
As AI becomes increasingly conversational, perhaps these seemingly small contextual layers will matter more.
AI-Watch is my experiment with one of them.
It's free, and I'd genuinely be interested in feedback from developers and AI users.
Most of all, I'm curious about one thing:
Do you think temporal context can meaningfully improve an AI conversation — or is knowing the current time on demand already enough?