A little while ago, I started using Claude more than I did any other AI model. It was a pretty huge moment for me, given how Google Gemini was my previous go-to LLM for all of my needs, but I found Claude performed some tasks way better than anything else. So, I decided to take the plunge and use Claude as my go-to AI to see where it excelled and where I'd be better off using a different AI.

One of the features I quickly fell in love with was Claude's memory tool. At first, I enabled it so Claude could get an idea as to who I am and better tailor its responses to me. However, the more I used it, the more I noticed Claude was telling me things about myself that I didn't even know about.

Claude's memory feature is very powerful

I didn't expect it to be this good

In case you haven't tried it before, Claude has an optional memory feature. If you enable it, Claude begins doing daily round-ups of all the things you discussed with it during and notes them down in its memory. The more you use it, the more it notes down. The more it notes down, the more it knows about you, your needs, and the kind of person you are. That way, you don't need to start every conversation with "Hey Claude, I'm XYZ;" you just tell it who you are, and it will carry that context into future talks

The best part about this feature is that you can actually take a look at what Claude remembers about you. You just open the settings page, pop open its memory file, and see its little "user diary" that it wrote about you. If Claude gets something wrong, you can tell it to correct something in its memory, and it will. Likewise, if you and Claude touch on a subject or idea you want it to remember, you can tell it to store it in its memory banks, and it will.

When I first enabled Claude's memory tool, I expected the AI to store some very basic, bullet-pointed facts about me. You know, stuff like my name, predicted age bracket, interests, etc; nothing more complex than a dating profile. However, the more I used the AI, the more fleshed-out the memory file got, until it was strangely detailed and in-depth. The best way I could describe the feeling is as if I were reading a personal evaluation of myself, including who I am, what I like, and how I act.

Claude turns your idle questions and prompts into a reflection of yourself

It built a picture of myself that I didn't even know I had

So, here's the thing. I'm not trying to claim that Claude is some kind of extremely intelligent computer that can analyse someone like a psychiatrist. When LLMs remember information about you and respond to you in a fitting tone, they're essentially mirroring you and how you acted. They're not taking what you put into it and dwelling on it to create a picture of who you are; they only know what you feed into it, and they often play off of the information you provide them to ask you more questions.

However, that doesn't mean Claude can't teach you a little bit about yourself that you didn't know. After all, its memory feature focuses on remembering stuff that people don't really give you feedback on in real life. You have bosses to give you an evaluation on your work, and your friends and family can teach you what kind of person you are, what you like, and what you dislike. However, LLMs can remember every little prompt you give them, from your curious questions to your full-scale projects, and use that to build a profile about you. Again, that profile is, essentially, just a mirror of you, but that doesn't mean it can't highlight things you've never seen in yourself before.

When I open my own memory file, I essentially see every prompt and discussion I've ever given it condensed into a 'Simon report'. Some of the prompts I don't even remember making until I read the report, but Claude remembered. And because Claude remembered every single prompt I made, it could tell me something about myself that I was unaware of. For example, here's what Claude says about me:

His workflow is iterative and accuracy-focused, with a strong preference for technically precise and defensible language over punchy-but-inaccurate phrasing. He reads primary sources and is receptive to corrections.

[...]

He has a pattern of intellectual curiosity across diverse topics — linguistics, legal edge cases, neuroscience, geopolitics, sleep science — and engages analytically, building understanding through follow-up questions and connecting new information to personal experience.

The rest of the memory file covers my personal and professional life, offering me insights as to how I handle challenges, how I operate best, and what I'm like to work with. This is all data that I have deliberately fed it, but Claude is finding patterns in my behavior that I have never personally recognised within myself. And that's really cool.

Claude remembers the bits that you forget

It's because Claude takes note of every little thing you do, from your serious work crunch to your casual question asking, that allows it to draw up such a vivid report on the kind of person you are. And honestly, reading an evaluation that an AI made of you can teach you something that other people just cannot replicate.