Almost no memory I have involves images. I have a few, but most things I remember are feelings or information that can't easily be described with words. If you tell me "remember that blond girl yesterday" I'd have no idea whom you're talking about. I've even realized mid-conversation that some of my former employees from a decade ago were black, but I didn't know who that was until I was told their names. My brain simply doesn't store visual appearance as a retrieval key.

This has a name: aphantasia — the inability to voluntarily form mental images. I scored 19 out of 160 on the VVIQ (well below the 32 threshold), which surprised even me — I had assumed I was merely low-imagery, but it turns out the mind's eye is almost entirely dark. You can measure where you fall on the spectrum with the VVIQ2 test

If you ask me to imagine a red apple, I can somehow sense the stem on top, the bottom indent. But I don't see it. Organic shapes collapse into abstract spatial knowledge. Technical objects — a computer, a house, a lock mechanism — I can reason about freely in my mind, but they appear as wireframes: structure without texture, if they appear at all.

Yet my spatial reasoning is strong. I can mentally float around every house or apartment I ever lived in—placing rooms relative to other rooms across different floors, locating windows, switches, furniture. If you've ever used a 3D level editor like Hammer (the one that shipped with Half-Life), imagine that: brush geometry, entities, relationships. That's roughly what my internal representation looks like.

Complex systems appear in my mind as something between a spatial diagram and a logical tree. Nodes and connections, dependencies and constraints—but the graph can't be forced into 3D space because relationships cross and overlap in ways that a physical layout can't represent. When I think about software architecture or infrastructure, I'm navigating that structure.

My memories are frozen instants I can inspect—like breakpoints in a running program. They're not video playback. Each one is a single frame pinned to a location on a spatial map, with some metadata attached: who was there, what happened, the emotional state. But no continuous timeline, no before or after.

Time itself is abstract for me. I can map spaces but not changes in space. I can update the map, but previous versions are usually lost. I have some memories locked in time because they're attached to a birthday or a known event, but generally I can't place memories on a timeline. Remembering what I did yesterday or last week is almost impossible without looking at pictures, unless something notable happened.

I now offload temporal indexing to technology—mostly Google Photos, where I can quickly search for objects or places and retrieve timestamps. It's essentially an external memory system: the brain holds the spatial map and the structural knowledge, and the phone provides the when.

As a child I would dismantle calculators, remote controllers, and anything I wanted to understand. More often than not I was met with a boring black epoxy blob hiding all the interesting parts. But the drive was always the same: see the mechanism, build the internal model, predict the behavior.

The flip side is that by wanting to understand and do everything myself, everything becomes difficult. I once wrote that the geek is a kind of Shadok: why do it simply when you can make it complicated? You'll spend all night compiling Firefox from source—patching, debugging, chasing dependencies—and at 6am, when the sun comes up, you apt-get install the release build and go to sleep. The point was never the browser. The point was the hundred things you learned along the way. This eventually led me to build Azusa, a full pseudo-Linux distribution where most GNU components — glibc and all — live under a dedicated directory, designed to work alongside another distribution. That means compiling everything myself, some packages far more painful than others.

I now know how the internals of a CPU work—from instructions to registers to logic gates to transistors—and I have a good idea of how a calculator or remote controller chip is made without dismantling anything. The same applies to anything mechanical: door handle mechanisms, locks, machines. I can usually imagine how something was made just by looking at whatever moving parts are visible, and quickly reverse-engineer the internal workings.

The same intuition applies to software. I can see inside most systems almost instantly, getting a good idea of the process used to reach the final product. I learned to content myself with understanding the process rather than being able to reproduce it fully. The reward is the model, not the object.

Debugging complex problems works best in the background. I keep them in mind, and the best place for solutions to appear is usually while taking a shower. The brain builds a complex relational model, and when I relax, the default network reorganises it until the correct structure emerges.