Memory-aware living | Casprine Assempah

I’ve never been good at remembering things. I can’t really pinpoint when it started but I’ve mostly moved through life with a fuzzy memory on good days and told by those who are close to me that I’m forgetful. This has caused some problems for me. Some say it’s trauma, I say we are not molded from the same clay.

I haven’t been medically diagnosed with a memory issue, mostly because I haven’t gotten tested. Maybe I should.

A large majority of my recollectable childhood memories are rooted in pain; arguing and fighting with my siblings to the point of almost poking one of my eyeballs out during a fight. I might have been a troublesome kid at some point. Things like this, I would gladly pay good money to forget in exchange for remembering the faces and voices of past family members.

A couple of months ago, I lost 2 older family members relatively close to me. The first few days after finding out were coupled with shock and general sadness at not being around them during their final times. My last physical interaction with either of them was in 2024.

Slowly, I began trying to recollect their faces, voices and the memories we shared together. I didn’t care whether they were pleasant or not, I just wanted to remember them.

I hit a wall. I really couldn’t remember much. I remembered their faces sure, but I could barely remember anything with them. My first course of action was to find something to trigger my memory, going through 5-year-old photos and scrolling through the family group chat to find a voice or video recording of them. This was a futile attempt. I couldn’t find anything substantial. Either a lack of photos and videos taken with them or just general negligence on my part.

Either way, I decided I should take steps to make this less likely to happen, hence tackling the grander problem at hand: not remembering things. The goal is simple; take a more intentional approach to remembering everyday life, one form of media at a time. This documents how I plan to make and store such media.

Journalling

Thoughts, emotions and general day-to-day events are collected and stored as journal entries. I write at least twice a day, when I sit behind my workstation to start work and usually around 18:00 before I get up to go make dinner. I also have sporadic mini-sessions during work whenever anything of importance comes up.

I started off with pen and paper. The tried and tested method. I loved it for the most part, but I can’t search it, paste it into Claude to give me a summary of the week and I have to actually store them somewhere. This is a hassle.

The first alternative was to go semi-digital. Write on paper, take a photo and use the OCR on the iPhone Photos app to select the text and paste it into Notes. Unfortunately, I have to manually correct some of the text. Too much work, I’m too lazy.

I’ve tried a lot of note-taking apps. Notion was the first. It was fast, easy, worked on even the shittiest of connections. Then it got bloated and I became a data and privacy nut, all thanks to David. Obsidian is the best one out there in my opinion, but it still didn’t solve everything for me. I want local-first, sync and easy AI features. I am hard to please.

So like every engineer before me, I decided to make my own. Just mine, how I like it, how I want it to work, look and what AI features I want, all the while keeping my data on my homelab and accessing it via Tailscale when I’m not home.

I started working on Aether last year and it’s been good to me. Still no AI features yet but it satisfies all my needs. I recently discovered WhisprFlow which makes it significantly easier to document daily changes and events using my voice.

Journal timeline view with activity heatmap

Walkthrough on creating goals & tasks

Media

My memories have generally been stored across multiple mediums and platforms, from iCloud to SSDs lying around the house. iCloud usually holds the most intimate ones, typically from my phone, and I kept my travel and camera photos on SSD drives.

For a while, I played around with the idea of a custom platform with David to store, manage and manipulate this media. A Dropbox-like alternative built to be hosted locally with backup solutions to satisfy the 3-2-1 backup rule. Unfortunately, because of time, work and the complexity of it, the project is currently on pause. Maybe with AI agents getting so good, I might actually build it to a point that’s usable.

The solution I settled on is a combination of a NAS (Network Attached Storage) and Immich as the software of choice for managing and previewing my media. The NAS is a SuperMicro SuperServer X10DDW-i running two Intel Xeon E5-2630V3 processors — 16 cores, 32 threads, 64GB ECC DDR4 running running TrueNAS. This server stores every digital file I own. Yes, I also scan my physical documents for processing. TrueNAS provides an easy-to-use interface and NFS to move files across the network. 1Gbps up/downstream is plenty enough for my needs at the moment.

Immich is like Google Photos, only better in the sense that it runs purely on your own hardware and you own your data. Geotagging, people recognition and all the nice features that Google Photos is known for are also available.

I’m still yet to figure out a simpler solution for backups. I’m currently looking into Backblaze and S3 as the cloud services of choice. I would have preferred to have it on iCloud storage but I haven’t found a way to get it working yet. If you have, please reach out.

Rewind

I spend most of my day on a computer; what I see, what I read, links I find. There’s a lot of context here that’s lost by end of day and in my case, forgotten. I was over the moon when I first read the release announcement of Rewind.

Advertised as the all-seeing, all-hearing personal companion that has all the context needed to help you remember. That’s the general gist of what it does. Prior to the AI boom, they didn’t even call themselves an AI company. The Rewind app records your screen and audio while you use your Mac, stores the data locally and allows you to go back in time with their timeline system. They also allow you to search for any word seen or heard and summarize your meetings, among other things. They deliver on their promise really well and the UI is quite delightful to use.

My only issue was the size it builds up after a while. I had roughly 100GB of my SSD filled up with their data after using it for a couple of months. It’s understandable considering the quality of the screen recording depends on how frequently you use it. This started causing my workstation to slow down.

The solution for that was to offload the excess recorded data to the NAS and build a renderer on top of it to view the content. Ideally this would be automatic, uploading the data on a schedule, twice a month. The hard part would be building the renderer to perform search and summarization, but one thing at a time, the locally saved data is encrypted anyway.

Unfortunately, Rewind’s parent company was acquired by Meta. Even though they said they are not going to train their AI on user data, no one trusts Meta with anything. They also announced they will be sunsetting the service. This is a hard one for me because I really found the app very useful. I’ve seen a few alternatives but none are up to the quality of Rewind.

I’m going to throw the problem at Claude and Codex to solve it. At least a rudimentary version that lets me always record my screen, do OCR on the content, store it locally, go back in time to see what I was doing at a particular time and also solve the storage size issue. I’ve already made a plan with Claude and will probably let the agents build it. I need this back in my life.

None of this fixes the underlying issue, but it gets me closer to a life where forgetting is harder to do by accident