An AI Assistant That Remembers, With Notes You Can Edit

You have explained your situation to a chat assistant three times this month, and it still starts from zero. Or the opposite: you open its settings and find it kept things you never meant it to.

Both are the same problem. An AI that remembers is only as good as your view of what it remembers. Ranya keeps short notes about the stable things you tell her, brings them into every chat, every call and your daily brief, and lists every one of them in Settings, where you can read, edit or delete it.

Why does my AI assistant forget what I told it?

A chat assistant works from the conversation in front of it. Inside one chat it can follow along: you said you work from home on Fridays, so it plans around that. Start a new chat and the page is blank again.

Your job, your week, the race you are training for, the way you like answers: none of it carries over unless the assistant keeps notes between conversations. So you explain again. Each time it is another paragraph of typing, and after a while you stop bothering. The answers get more generic, and the assistant starts to feel like a search box.

Some assistants now keep a memory between conversations. That fixes the repetition, and opens a second question: what exactly did it keep, and where can you see it?

Why should an assistant's memory be visible?

Because a memory you cannot see turns into a surprise. In a 2026 study, researchers interviewed 20 ChatGPT users about its memory feature, and this is what most of them went through.

They described a strong need for more visibility and more control in future memory features. And the wish for control is wider than memory. Asked about AI in their own lives, most Americans want more say.

That is the brief for an assistant that remembers: keep what helps, and put all of it in plain sight, where you can read it, fix it and delete it.

What should an AI that remembers keep about you?

The useful things are the ones that stay true for months: how you like answers, when you work, what you are training for. They change how an assistant should help you, and you should not have to say them twice.

Plenty of what you say is not like that. "Call the dentist tomorrow" is a task, and it belongs on your list with a date. "Read 2 books" is a goal you can check. "Make this reply shorter" is about this reply, not about you.

Ranya is built to draw that same line. She keeps the stable things as short notes, one sentence each, filed under a topic such as Work, Health or Preferences. One-off requests, and what you already track as tasks, goals and habits, stay where they are. She is also built to leave out private details, such as medical information, money or passwords, unless you ask her to keep them.

Does Ranya save things on its own?

Yes. When you mention something stable, she saves it as a note without asking first, and says so in a few words. It is one of the very few things she does without your yes. Everything that changes your work, such as a task, a habit log or an email, is read back first and waits for you. Here is why Ranya asks before she changes anything.

Why build memory the other way round? A question every time you mention a preference would turn a conversation into paperwork, and you would learn to stop mentioning things. A note is small, you can see it as soon as it exists, and you can change or delete it whenever you like. So the trade is simple: she saves on her own, and you can always see what she saved.

In chat you watch it happen. A small Remember line appears in her reply. Open it to see the note and the topic it went under.

Where does Ranya use what it remembers?

In three places. Every chat starts with your notes, so when you ask about your week, the fact that you work from home on Fridays is already part of the conversation. Every call starts with them too, so when you plan your day out loud on a call, you skip the part where you explain your situation.

And your daily brief is written from your own logs and check-ins, your why, and what Ranya remembers, so a note about the race can shape what it suggests for your training. She is built to use notes as background, not to recite them. See what else you can do on a call with Ranya.

How to see, edit or delete what Ranya remembers

1. Tell her something that will stay true

   Say it or type it plainly: "Remember I'm training for a half marathon in October." A stable fact is usually enough on its own, but saying remember leaves no doubt.

2. Look for the note

   In chat, the Remember line in her reply holds the note and its topic. On a call, she tells you she has it.

3. Open Settings, then Memory

   Or say "open my memory settings". Every note is there, grouped by topic. You can write your own too: New topic, or the plus on any topic.

4. Change a note when life changes

   On a computer, point at a note and choose Edit memory or Forget memory. On a phone, tap a note to edit it, or swipe it left to forget it. You can also say "I work from home on Thursdays now, update that", and she reads the change back before she saves it. Forget everything clears the whole list, after one more question.

5. Say "forget that"

   On a call or in chat, she finds the note and asks before she deletes it. On a call the card says Delete, in red; in chat the button says Forget. Say no, and the note stays.

   Ranya

   2:05

   Ranya is asking

   You: Forget the half marathon, I dropped out.

   Ranya: Forget your half marathon note?

   Ranya is asking: Forget

   Ranya is asking

   Forget

   CancelDelete

   MuteEndMinimise

   Example call on a phone. You say: Forget the half marathon, I dropped out. Ranya asks: Forget your half marathon note? The card reads Ranya is asking, Forget, with Cancel and a red Delete button.

6. Notice where it helps

   Your next chat, your next call and tomorrow's brief all start with the notes you kept.

How to keep an AI's memory useful

• One fact per note. Short sentences are easy to scan and easy to fix.
• Update, do not pile up. When something changes, change the old note, so two notes never disagree.
• Keep tasks as tasks. A to-do in memory has no date and no reminder. On your list it has both, and you can let Ranya keep the list with you.
• Rate the brief. Each goal's card in your daily brief can add one thing worth knowing, which it calls the spark. When it asks "Was the spark useful?", a thumbs up or down is saved as a note under Preferences.
• Read the list each season. Forget what is no longer true.