An assistant that remembers to ask how it went

It is 6:17 on Saturday evening. You open the laptop to find a receipt, and a small speech bubble appears beside Mia: How did Friday’s presentation go?

Not “What can I help you with?” Not another summary of your calendar. The thing you were worried about three days ago has happened, and the assistant remembers that the story had an ending.

That is a tiny interaction. It is also one of the clearest differences between an assistant that answers questions and one that shares a little continuity with you.

Most conversations stop too early

Suppose you tell an ordinary chatbot, “I have my final interview on Thursday.” It may help you prepare. It may produce good questions, polish an answer, and wish you luck. Then the conversation ends. On Friday, the interview exists only as old text you could search for yourself.

A thoughtful colleague would ask what happened. A friend would remember the flight, the appointment, the parent-teacher meeting, or the conversation you had been avoiding. The follow-up does not solve the event. It shows that the event was not treated as disposable context.

This matters most to people whose work and life contain many open loops: job seekers moving between interviews, managers preparing for difficult conversations, freelancers waiting on pitches, students facing exams, and anyone trying to keep track of a week that refuses to stay tidy.

The feature begins with something you actually said

DT Life looks for a future event in the messages you send. The timing must be present in your own words—“on Thursday,” “tomorrow morning,” or another real future reference. If the message describes something worth asking about, the app saves one follow-up for a sensible time after it should have happened.

A calendar event that has just ended can create the same kind of question. Both sources enter one queue, so mentioning the same plan repeatedly should not produce repeated questions. When you answer, the answer closes the loop.

The important part is what DT Life does not do. It does not turn every noun into a life event. It checks that the words supporting the time really appeared in your message, rejects backward-looking references, and will not schedule a question more than 60 days away. Those boundaries exist because a fluent local model can still mistake a memory for a plan or invent a plausible date.

When the moment arrives, Mia uses the same reminder system as the morning briefing and meeting preparation. The question appears in a companion bubble and can be spoken if auto-speak is enabled. Focus time can hold it until a better moment. Pending questions are visible in Talk → Diary, where you can drop one before it is asked. A switch in Settings pauses follow-ups, and the main reminders switch turns all proactive moments off.

Remembering should not require an audience

A follow-up can reveal more than the original sentence. “How did the oncology appointment go?” or “Did you tell your manager you are leaving?” is sensitive even without the answer.

DT Life keeps the message, the planned question, and any answer in its local database. Private text is encrypted at rest with a key sealed to your Windows account, and the wording is generated by the model on your machine. There is no AI service receiving a list of the events that matter to you.

That privacy is not an extra badge on this feature. It is what makes the feature reasonable. An assistant cannot build continuity without remembering personal things; the user should not have to trade ownership of those things for the convenience of being remembered.

It will sometimes stay silent

If you write “I might do it sometime,” there is no dependable time to use. If the calendar title is vague, the resulting question may be vague too. A small local model can miss a valid plan, and it is better for the feature to miss one than to confidently ask about an event you never had. Reminders also need to be enabled for Mia to speak up.

The goal is not to imitate perfect human memory. It is narrower: when you tell your assistant that something important is about to happen, that fact should not disappear just because the chat window closed.

See the companion features →