The Render Runs At Night My mind runs on MIDI

A friend told me, the way you’d mention the weather, that there is no voice in their head. No narration. No little announcer reading their thoughts back to them as they think. I asked three follow-up questions in a row because I was sure I’d misheard, and they answered all three with the patient confusion of someone being asked to describe the color of oxygen: there was just nothing there. Thinking happened. It simply didn’t talk.

I’d spent that same week reading about aphantasia – people who can’t summon a mental image, who close their eyes and get an honest black (or rather, Eigengrau) nothing where the rest of us apparently run a slideshow. I’d filed it under fun neuro-trivia and moved on. Then the no-voice thing landed, and the floor tilted a little, because it cracked an assumption I’d been standing on my entire life without ever noticing it was load-bearing: that everyone boots the same operating system. That the inside of my skull is a standard-issue room and I’ve just been describing furniture everyone already owns.

Turns out there’s a word for my friend now: Anendophasia – coined in 2024, for exactly this, a life with no inner speech at all. And what got me wasn’t the diagnosis – but realizing I’m the photo negative of it. Whatever the maxed-out far end of that dial is, the setting stuck on eleven, that’s where I live.

Because I don’t have an inner voice. I have an inner cast.

There’s a voice, sure. But there’s also a second one that immediately cross-examines the first, and if I’m actually chewing on something I can convene a whole table of them – a skeptic, an enthusiast, the version of me that’s still annoyed about something unrelated from Sunday – and let them argue it out while I watch from a fixed point above the table, like a camera on a boom. Charles Fernyhough calls the back-and-forth kind dialogic inner speech, and the genuinely funny part is that arguing with yourself like this seems to lean on some of the same machinery your brain uses to model other people. Which would make my little parliament of selves a theory-of-mind trick pointed inward. The whole council is just me, cosplaying as everyone I’ve ever met.

And underneath all of it, always, a soundtrack. Instrumental, never lyrics, running like a radio station nobody staffed. I assumed, the way you assume gravity, that everyone had this going on. Turns out the friend with the silent head thinks I’m the strange one, and for once I can’t argue.

So I tried to picture a face

For years I had a tidy theory about all this. Obviously I was a visualizer: big inner world, cinematic camera, a whole repertory company in my head – surely that ships with a big screen, right? Full mental IMAX.

So test it. Close your eyes and picture the face of the person you see every single day. The one you’d know in a crowd from the back of the neck.

I did. And what came up was not a face, and not a blank either. It was a nebula. A structureless cloud of glowing dots at different brightnesses, in a specific color – jade green, glittering, shading light to dark – wrapped in a whole weather-system of emotion and memory. That is my partner. Not a picture of him. A color-field with him encoded in it.

And if I go looking for the actual face, I can retrieve the parts. The snub nose. The accentuated black brows. The dark, short hair, the glasses. The exact way his eyes move when they’re playing with a room. The texture of his skin under my hand, the smell of him – all present, all real, all individually there. What I cannot do is assemble them. I can inspect any single feature and I know precisely how it connects to the ones around it, and I still cannot build the whole face out of the parts. I could describe him to you piece by piece and then fail, completely, to draw him. I have every component and the entire wiring diagram, and the one step that would composite them into a thing I could actually see never runs.

That should have been unsettling and instead it was thrilling, because it detonated the visualizer theory on the spot. You cannot be a maxed-out super-visualizer with a nebula where the most important face in your life is supposed to be. The face was never the file – the file is the nebula. The face is just the one export format my renderer doesn’t support.

And it’s not only him. Everyone I know arrives as their own palette and their own nebula – a personal color scheme, no two alike – and the palettes drift over time, updating as the person does. That is simply what people look like on the inside of my head. Where the colors come from, why his is jade and someone else’s isn’t, I genuinely don’t know. I haven’t traced any of it back to a source. Call it an open question I haven’t yet put to the right members of the inner parliament.

Positions with attributes

So: not a visualizer with a bug. Something else entirely.

Turns out mental imagery isn’t one thing. There’s object imagery – pictures, color, texture, the glossy photographic kind – and there’s spatial imagery, which is relationships, positions, how things sit and move relative to each other. Two different systems. And the research on it found they don’t just differ, they tend to trade off – strong on one often means thin on the other, as if the brain rationed one pool of resources and made you pick. Visual artists cluster at the object end. Scientists and engineers cluster at the spatial end.

Means – I am a frontend engineer who thinks in graphs.

When my sister was over recently we ended up playing one of those dumb long-drive games where you narrate how your own head works, and somewhere in it I said, without thinking about it at all, I only see positions with attributes. Then I actually heard the sentence come out of me. Because that is not a description of a broken visual system. That is a data structure. A node with properties hanging off it – and now I could see that the nebula was exactly that, a set of positions carrying color and brightness and emotion as their attributes. I don’t store the scene; I store the graph of the scene and let the meaning hang off the nodes.

This is where the tidy population story wants to claim me as its poster child – the object-imager kid who grew up into a spatial-imager adult, one pole traded clean for the other. Except that’s not what happened, and I’ll get to why at the end. I didn’t trade one for the other. I kept both. The between-subjects average makes a lovely arc and it is not my biography; hold that thought.

God’s-eye, and clumsy

Because the spatial half isn’t a metaphor. I’m Google Earth in there.

Drop me somewhere I’ve never been and I orient almost instantly. Whatever I see once, I keep – for years. Jayden and I take the motorbike out on long trips and I check the route on the map exactly once, before we leave, and then I have the whole thing, both directions, including alternative routes, and never look at the phone again. When I’m about to visit a new city I explore it first from above on Google Earth (I spend an indecent amount of time up there, prowling the corners of the planet), and when I land I can already move through the streets, because I’ve seen them from the top – the landmarks, the street patterns, the texture of the ground. Ask me to hold a labyrinth in my head and rotate it, mirror it, alter it – easy, every angle. I could build you my company’s office in 3D from memory, and those of all previous companies, every room and wall and desk roughly where it belongs.

And I bump into strangers on the pavement.

That’s the part that turns this from a flex into something truer: I am superb at the map and clumsy in the body. God’s-eye view: flawless. Where-am-I-right-here, relative to the thing within arm’s reach: weak. I genuinely lack a strong sense of my own body’s position, which is why I misjudge the gap and clip someone’s shoulder, why an obstacle in a corridor with two perfectly good ways around it can strand me in the doorway while I fail to just pick one. I can model the whole city from above and I can’t reliably get my own coffee cup to my mouth without incident.

Which is the same fact twice, not two facts. The vantage I model the world from is external – above it, outside it, a camera on the scene rather than a pair of eyes inside it. There’s real work on this, actually: Nigro and Neisser named the two ways people replay their own lives – through their own eyes, or from outside watching themselves – and the outside-watching kind turns out to be more common for childhood and for anything deep in the past. For me it isn’t a mode I switch into. It’s the only camera I’ve got. I see the office from above because the overhead shot is the native view. The price of living at the god’s-eye vantage is that there’s nobody home behind my actual eyes to reach for the cup.

Keep that camera in mind, too. It comes back, and it never once moves.

It was never about seeing

Once the visual thing cracked open I couldn’t stop pulling the thread, and the thread kept coming, because it was never really about seeing at all.

Try to remember a conversation. Not what was said – the sound of it. The actual timbre of the other person’s voice, the specific grain of it. I mostly can’t. I remember what was said with high fidelity, the meaning and the weight and who came out on top. But the audio isn’t faded, it’s absent. It was never rendered in the first place, so this isn’t a visual gap. It’s the whole sensory surface, dropped across every channel at once, picture and sound alike, while the meaning and the emotion and the relational structure stay pinned at full resolution. I keep the wave function and throw away the pixels.

With a handful of exceptions, and the exceptions are the tell.

There’s a small set of memories that kept their sound – intact, recall-on-trigger, not on loop. Some carry grief and plenty carry the opposite: things said in love, in joy, the good ones, held in the actual voice that said them. What they share isn’t sadness, but intensity. Turn the emotional dial high enough, any direction, and the moment writes itself to disk with the audio still attached, when every ordinary moment arrives as a silent transcript.

The clearest one I have, I’ve written about before. Six years ago now, I flew in from Vietnam and finally made it to my mother’s hospital bed, and I told her there were so many things I’d always meant to say and never said, things I’d wanted to show her and never did. And she answered – just one sentence: Wir hatten keine Zeit mehr. We had no time left. It was near the last of her strength, and it was the very last thing she ever said to me. She was gone a few days later.

I can’t hear my closest friend’s laugh. But I can hear that.

Six years is enough time to make peace, and I have – what’s left is sadness, not grief, and letting people go turns out to be part of the deal rather than a rupture in it. So I’m not telling you this from inside a wound. I’m telling you because it’s the cleanest evidence I own that the renderer isn’t gone, just off, and that enough force can reach in and flip it on for exactly one moment and burn the thing in permanently. I’ll add the caveat the rest of this essay demands: I can’t swear that recording is accurate to the letter. Brains re-process and defragment and rewrite in the dark, and mine is subjective like everyone’s. It’s the closest thing to a true recording I’m aware of holding, which is not the same as true.

The music is a MIDI file

Okay but… the music. I’ve been sitting on this the whole time, because it looks like the exception that sinks the theory. If sound is stripped out of everything, how do I have a full orchestra playing in my skull right now, and how has it been playing my entire life?

So I tested it, with some the most reliable soundtracks I own: the Vietnam theme from Civilization 6’s atomic age, for example. And as long as I don’t turn my attention onto it, it plays. It feels like music – the melody’s there, complete, in the right tempo, and I can whistle it out loud straight off the internal version. I can speed it up and slow it down at will.

But I cannot change its volume. And the moment I actually focus on it – bring the background process into the foreground and look at it directly – the music isn’t there. What’s there is the MIDI-like data underneath it, the instructions, and some translator in between that was rendering those instructions into the illusion of sound the whole time I wasn’t looking. Is there a melody in my head? Yes. As long as I don’t listen to it.

That volume detail is the giveaway. Tempo is an instruction – a number in the file, so of course I can change it. Volume is a property of actual sound, and there is no actual sound to make louder, so I can’t. The music is the score, not the performance. And this is the whole architecture in a single object: everywhere, across every channel, I hold the instructions and skip the render, and the render only runs while I’m not watching it. Same trick as the face, which is a parts-list I can’t composite. Same trick as memory, a transcript with the audio stripped. The music just happens to be the one place I can catch the render running and dissolve it on command, live, by the simple act of paying attention. Look at it and it collapses into data. The observer’s barb, aimed at a song.

Which, annoyingly, is also why it wrecks my sleep. There’s a sleep study on bedtime earworms dragging down measured sleep quality, and the cruel twist in it is that instrumental music, the kind with no words, lodges worse than lyrical. My internal station has never once played a song with lyrics. And earworms thrive in exactly one condition: low attention. Which is precisely the state I’m in at 2am, drifting, not looking at anything – so the rendered version, the one that only runs unwatched, is running at full tilt right when I’m trying to go under.

I may have built the station myself, too. As a kid I fell asleep to all sorts of game soundtracks – Settlers 2, the Age of Empires 1 score, the ambient city-hum of an old mob game called Gangsters, and a Christmas-gift CD of pseudo-Egyptian orchestral music that I much later realized had trained me, straight Pavlov, to drop unconscious the second I heard it. Play me the right track today and I still get heavy-eyed on cue. I wired the thing directly into the sleep switch, one bedtime at a time, and then it refused to ever power down.

The good graphics are at night

Where it inverts, and where I almost told you the wrong story: awake, I render almost nothing. Asleep, I render everything – full color, full sound, full intensity, every single night, lucid and persistent, with recurring dream-worlds I return to across years like reopening a save file. The obvious conclusion, the flattering one, is that I’ve got a secret cinema in there that only unlocks at night, a hidden renderer sleep sets free.

Except that’s not what the evidence says, and the evidence is sitting right in the dreams themselves.

The people in my dreams don’t match the real ones. My dream-mother, my dream-friends – they’re vivid, more vivid than anything I can summon awake, but they are not accurate likenesses. They’re what my imagination invents as their appearance, built from the semi-structural data I actually store: the attributes, the emotion, the sensory scraps. If there were a hidden photographic renderer in there holding the real faces, the dreams would show me the real faces. They don’t. They show me plausible fabrications assembled from the parts-list. Which means there is no second, better engine waking up at night. There’s one engine, the same one, running the same attribute-file it always runs.

What sleep actually does is lift two governors off it. It takes the cap off the volume – everything gets turned up, brighter and louder and more intense than waking imagination allows. And it releases the brake on compositing – the assembler that flat refuses to fire while I’m awake, the one that leaves me holding a face I can’t build, finally runs. But it composites from the data, not from a photograph, which is exactly why the dream-people come out wrong. The engine is a superb artist and a terrible photocopier. Give it the attribute-file and it will paint you something breathtaking; ask it to reproduce a real face and it improvises one instead, and never notices the difference.

(The rest of what happens down there – the unpinned self, the body that comes back a different age or a different gender or occasionally not a person at all, the register that runs gory and apocalyptic as easily as tender – is real, and it’s genuinely its own essay, so I’ll leave it gestured at rather than itemized. The short version: the part of me that pins down this is who you are, hold still is exactly the part sleep switches off, and once it’s off, the self stops being one fixed thing. For a pansexual guy who’s never felt especially nailed-down to begin with, that lands as accurate rather than alarming. More another time).

And through all of it – the one thing that never moves – the camera. Third-person or even cinematic camera, outside my own body, watching, same vantage as my waking memories, and more often than not I’m a kid in there. The volume changes, the compositing changes, the self comes unpinned. The camera stays exactly where it always is. It’s the one fixed point in the entire system.

So: not a caged renderer. A braked one. The executive-control story – that some top-down grip holds the whole thing offline while I’m awake – is my best guess at the wiring, not a reading off a dial. But the split it’s guessing at is real, and it isn’t even rare. Summoning an image on purpose and having one arrive unbidden turn out to be two different machines: plenty of people with no waking mind’s-eye at all still dream in pictures. The deliberate renderer and the automatic one come apart cleanly enough that you can be dark on one and lit on the other. What might be unusual about me is only how far apart the two sit – nothing on command, a whole cosmos at night. And even that is a thing I feel, not a thing I’ve measured.

Which is why mornings are like that

Most people wake up out of a dim, half-formed dream into a high-resolution world. They climb the fidelity gradient the instant their eyes open. Reality is the sharpest thing they’ll touch all day.

I go the other way, and I want to be precise about how, because it’s easy to overstate: it isn’t that waking reality is low-resolution for me – the room is as sharp as it is for anyone, my eyes are perfectly good. It’s the inner track that drops out. I leave a full-render, full-color, surround-sound universe and arrive in a waking mind that renders none of it back to me, and the gap between what I just left and what my imagination will now give me is a cliff. I wake up with the world in perfect focus and the mind’s eye behind it switched off, and that specific drop is a flavor of disillusionment I get most mornings. Not depression. A gradient. Homesick, at 7am, for the only room where my own renderer is allowed to run at full brightness.

Don’t look directly at it

There’s a related trick this whole setup can pull, and it comes with a warning taped to it.

Run several thoughts at once – I do, more or less always – and you hit a hard bottleneck at the mouth (or the typing fingers), because speech is single-file and thought is not. I stumble over my own sentences fairly often, and as a frame for it, not a diagnosis: it’s rarely that I lost the thread, it’s that three of them arrived at the exit together and the conclusion had already walked off before my mouth finished the setup. Parallel process, serial output bus. And it isn’t only speech – drop an obstacle in a corridor with two equally good ways around it and I can stall right there in the doorway, both routes valid, no tiebreaker firing, something awkward loading. Same bottleneck, one mouth and one body, both single-file.

But the sharp end of it is this: point conscious attention at something built to run on its own, and you can perturb it just by watching. There’s a name for the harmless version – the centipede effect, the bug that walks fine until it’s asked to explain how, at which point it trips over its own legs. Watch your own speech too closely and you stutter. Turn attention onto your own heartbeat and it starts to feel wrong, off-rhythm, like reaching for a stair that isn’t there in the dark – not because the attention is rewiring the organ, but because the noticing itself distorts what you feel. And in my case that one isn’t purely a party trick, because I’ve got a diagnosed quirk in the heart’s own electrical wiring, the real thing, quiet and medicated for a couple of years now, which gives the noticing something genuine to catch on.

Point the same beam somewhere higher-stakes and it stops being fun. Years back, around 2009, I spent a stretch trying to consciously observe and steer a perception that was already not behaving in the usual way – I’ll leave the specifics in the drawer where they belong – and it went about as well as staring down your own heartbeat and politely asking it to hold still. Same mechanism. Same barb. Much higher stakes. The observer perturbs the observed, and that’s just a fact about attention; it does not care whether you’re watching a sentence, a pulse, or your own grip on the floor.

A fair warning, this late in the tour: every word of this map was drawn by the one instrument in the building I can’t calibrate: me, watching me, and even watching me watching me.

The whole reason Hurlburt had to build a method – the random beeper, the careful interviews, the years of it – is that people are catastrophically bad at reporting their own insides without training. We confabulate. We round to the nearest tidy story. We describe the furniture we already expected to find and call it looking. Everything you just read is me doing exactly that, at length, with no beeper and no supervisor.

And it’s worse than untrustworthy, because you already have the tool to see why: the section right above says that watching a process bends it. So the act of writing this down changed the thing it was trying to describe. I can’t ever catch my own mind unobserved – the observing is the catch. The one system I’d give anything for a clean look at is the one permanently blocked from view by the fact that I’m the one looking. Instrument and experiment, welded together, no prying them apart. Ever.

So hold all of it loosely; this isn’t a diagram. It’s a sketch of a crime scene by the only witness, who is also the suspect, and who smudged the scene leaning in to look.

I drew it anyway. A smudged map of somewhere extraordinary still beats no map at all.

So where did it go?

Because that’s the question sitting under all the neuro-trivia, the one that actually sent me digging.

I was an art kid. Drawing constantly, painting, art school, the accordion, and a gigantic invented universe my sister and I built together over years, still alive in both our heads today. And somewhere in my teens the drawing tapered off, and for a long time I filed that under loss. The classic story – the talent that shows up early and then leaves.

Then I actually looked at it, and the loss story fell apart, because most of that analog output was never mine in the first place. It was my mother’s. The art classes, the instrument – every kid in the family had to play one – were scaffolding she put up around me, and it was good and warm when I was a child. Then puberty hit, around 1998 and 1999, and the enforced stuff started reading as uncool, the way it does for everyone, and at the exact same moment games walked in and opened a completely different door: editors allowing for worldbuilding, storytelling, and strategy, endless strategy. The scaffold came down and the thing it had been supporting didn’t collapse. It moved. Puberty and digitization retired the analog channel, and the engine picked up a new medium and kept going.

Because it never actually stopped. I make things constantly – images, music, web design, text, whole running stories with my sister and friends – and there’s a parallel world going the entire time in the background that re-renders every situation I’m in at a slightly higher setting. Not louder. More over-the-edge: more slapstick, more escalation, more feeling. It’s the thing that lets me accept the flatter real world, and occasionally it’s the thing that makes me lose it laughing at a private joke in a room where nothing funny is supposed to be happening.

So the creativity was never gated, and it never left. There was no guard, no clamp, no tragic switch thrown in adolescence. There was a change of medium I spent years misreading as a disappearance. The loss was an accounting error.

There’s a clean piece of science to end on, and it’s the one that reframes the whole thing: Russell Hurlburt spent decades interrupting people at random to ask what was actually in their heads, and found that a solid quarter of ordinary waking thought is what he calls unsymbolized thinking – definite, specific, fully-formed thoughts carrying no words and no images at all. So the way my own thoughts show up – total and weighty and completely shapeless, all meaning with the picture and the sound stripped off – isn’t a defect. It’s a documented, ordinary way of being a mind. I just happen to run in it almost full-time.

The engine has been running the whole time, in every medium I ever handed it – pencil, then an instrument I didn’t choose, then game-worlds, then code, then whatever this is right now. The only thing that ever changed was what it reached for. I kept mistaking that for silence.

It was never silent. I just wasn’t listening to it the right way.

Which, if you’ve read this far, you already knew I’d say.