LLM as a secretary

LLMs are good at understanding natural language, so a simple idea came up: why not use one as a secretary?

Not as a task manager or a control system. A good secretary doesn’t track progress or push you to complete things. Their job is to remind you when the time comes, in a calm and unobtrusive way.

You just tell it about something the way the thought appears.

In practice it works like this. You send a short message on the go: “remind me tomorrow evening”, “on Friday about the documents”, “in a couple of days”. You don’t have to choose dates or fill in forms.

The system understands the message, figures out the date and time, and sends you a reminder when it’s needed. It doesn’t ask you to confirm anything and doesn’t expect the task to be completed. It simply reminds you.

Because of that, there is much less tension. You don’t have to manage anything — you just rely on the reminder.

The interface is Telegram, because it’s already part of everyday communication. You can send a message the moment the thought appears, and the reminder comes back to the same place. It feels more like a conversation with a personal assistant than using a separate service.

You can try the bot here:
👉 https://lnkd.in/dA-nAyf6

On the technical side, the setup is quite straightforward. A backend built with Django and PostgreSQL handles data storage and scheduling. An LLM processes the text and extracts meaning, date, and time. Notifications are sent through a Telegram bot, and delayed reminders are handled by a task queue.

The result is a bot that understands loosely phrased messages and turns them into reminders. It doesn’t create a sense of obligation or pressure. It simply reminds you at the right moment.

That’s exactly how a digital secretary should work.