I use an AI as an external hard drive for my own memory, and the strange part is how much better my thinking got once I stopped asking my brain to store everything

I started using AI as an external memory. Somehow, it made me feel more human — which feels like it should be a bigger contradiction than it is, but here we are.

For the last couple of months, my life has felt like pure chaos. Not dramatic chaos, necessarily. Not the kind where one huge thing happens and everything collapses around it. More like the quieter, more exhausting kind, where there are too many open tabs in your actual life, and no matter how many you close, five more appear, smugly, like they’ve been waiting.

Work. Writing. Research. Personal matters. Challenges I give myself, because apparently the ones life hands out for free weren’t enough. Travel planning. More travel planning. Three different trips, somehow existing in my head at the same time, none of them fully booked. Apartment hunting. Life decisions. Decisions that are not technically urgent but feel existential anyway. And, underneath all of it, the familiar background anxiety of being 27 and still not having figured out one single stable thing about what I’m doing with my life — which is a very dignified way of saying I have a colour-coded calendar and no idea what I’m doing.

At some point, I told myself: give me a break.

But a break isn’t really possible when even a huge planner isn’t enough to gather all your daily quests.

I own several planners. Beautiful ones. I like the fantasy that if I buy the right notebook, my life will finally become legible.

Plot twist: it never does. It just becomes a very well-organised record of how ungovernable I am.

So I turned to the one thing I’m actually good at

So I realised I needed to use technology on myself, which, to be fair, is something I’m good at. I used Claude Cowork to store my own memory — quite literally.

Yesterday, I gathered all the latest things I have: online, offline, inside my laptop, inside my chats with ChatGPT and Claude, emails, even some messages I’ll admit to having kept, my own reflections, and the dreams I write down when I wake up and still, against all evidence, believe mean something.

I stored all of it — all my plans, all of me — on Claude. Literally, I stored myself in AI and created a Git repository for it. Private one, of course. But still.

A grown adult, version-controlling his own personality.

I created a living file that will get updated the next time something happens in my life. Something I’ll share with AI — and trust me, it will happen, because something always does, usually three things at once, usually while I’m trying to book a flight.

It feels mechanical. It’s also a little scary

Even though it feels mechanical to have an external drive for memory, and honestly a bit scary too, I feel exactly the way my Google Drive should feel whenever I clean it of unwanted files — which, if I’m honest with myself, happens about as often as I actually use those planners.

So it’s a good feeling. So far, at least. Ask me again after the AI has read my dream journal.

There’s something to that discomfort worth taking seriously, though, because I don’t think it’s just me being dramatic about a file structure. Memory feels like the most intimate thing we have — not just information, but texture, mood, the version of you that existed when something happened. Handing pieces of that to a piece of software felt, at first, like turning a life into files. Which, in fairness, is exactly what I did, on purpose, at 11 pm, with a slightly manic sense of accomplishment.

But then I realised something a little deflating: I had already done that. My life was already scattered across Google Docs, Notes app pages, screenshots, and documents named something like “new thoughts final final maybe.” I hadn’t been keeping my memory pure and untouched by technology, like some kind of monk. I had just been keeping it badly, and calling the badness “being a free spirit.”

Philosophers, unhelpfully, got here before I did.

In 1998, Andy Clark and David Chalmers proposed the extended mind thesis — the idea that cognition doesn’t stop at the skull, using the example of a man named Otto whose notebook does the remembering his brain can’t. My Git repository is not Otto’s notebook and I do not have his condition, but the underlying point held up better than my ego wanted it to: the mind was never as sealed off as I liked to imagine. It’s just that most people externalise their memory badly and unofficially, and I decided to do it badly and officially, with version history.

A container, not a replacement

I want to be clear that this is not the “AI will solve my life” way of thinking.

I don’t believe that.

Honestly, I find that kind of thinking a little dangerous, and also a little insulting to how hard my life has worked to be complicated.

I don’t want a machine deciding who I am, what I should do, or what my life means. I just needed a place to put things, because the place I was using — my own head, mostly out of stubbornness — was full.

There’s a psychological concept for this that I only looked up after the fact, which tells you everything about how this project actually started: transactive memory, described by Daniel Wegner in the 1980s from studying couples who stop each trying to remember everything and instead split the job — one remembers birthdays, the other remembers where the passports are. Apparently, I have now assigned this role to a chatbot, which either means I’ve optimised my life or given up on finding a partner who’s good with passports.

Possibly both.

What actually changed

The strange part is not that AI helped me remember more. The strange part is that once I stopped using my brain purely as storage, my actual thinking got better. Thinking, it turns out, is not the same as hoarding information, no matter how many years I spent operating as if it were.

When everything was inside my head at once, I was not necessarily thinking deeply. I was just drowning quietly in my own archive while presenting, outwardly, as a functioning person with opinions. There’s actual research behind this — cognitive offloading is the term for using an external tool to reduce the mental load of holding information yourself, and recent work modelling it as a value-based decision found people offload more exactly when the cost of holding everything themselves is highest — when memory load is greatest and internal storage is most strained. Reader, the cost was high. The cost had been high for months. I was simply too busy holding everything to notice.

Once things had somewhere to go, I could actually think about them. I noticed a worry from three weeks ago was the same worry, wearing a different costume. I noticed a decision I kept circling was one I had already half-made in a chat I’d forgotten existed.

None of that is the AI being clever. That’s just what happens when you’re no longer the only filing cabinet.

Of course, I still think this requires caution, and I say that mostly to reassure myself.

Privacy matters. Boundaries matter. Not every feeling needs to become data, and some things should probably stay messy, on principle, even for someone as committed to systems as I apparently am. But I am not using AI because I want to stop being a person.

I am using it because being a person has become strangely administratively demanding, and no one warned me that turning 27 would feel less like a milestone and more like being appointed, without consent, as my own project manager.

Maybe what I built is not a second brain — a phrase I’ve never liked, since it implies the self can be optimised into a clean, efficient duplicate. What I built is closer to a second memory room, somewhere I can leave a box, label it badly, and come back to later.

AI can hold the file. I’m still the one who has to recognise myself in it, dream journal and all.