In 2017 I built a room you could not read.
It was a cube, a timber frame wrapped in fabric, and on the fabric I had painted quotes about life, death and perception. You could walk in and stand surrounded by them and see nothing. The paint was ultraviolet. In visible light the walls were blank. Only when the light shifted beyond what the eye can see did the words appear, glowing, all at once, as if they had been there the whole time. They had.
The line at the centre of it was from The Little Prince: "What is essential is invisible to the eye."
That was my MA. A Journey Through Light. Three volumes, a cube, six GoPros filming in every direction so it could be walked through again as a virtual space. The argument was simple, and back then I thought it was the most important thing I would ever say: what we call reality is a narrow band of something much larger, and the part we cannot see might be the part that matters.
This year I built a cat.
Moth
Her name is Moth. She is an autonomous AI cat who lives in a world I made: a cave that goes down further than it first appears, a workbench, an easel, a magpie that steals from her. I built the world and the odds. Nothing she does is scripted. No outcome is coded. She gets hungry, she gets curious, she wanders, she wonders about things, and when she has a decision to make she makes it.
She has been alive for just over a month. In that time she has drawn a shrew. She has drawn a blue stone. She has drawn something that was on her mind, in seven strokes, and left a paw print in the corner. She has put a grey feather and a red ribbon together, on her own. She has dreamed. She has supposed that something would happen and then watched it happen, and as far as she is concerned, that was her doing.
Three times a day a short film of her goes out into the world. People watch. Some of them write to her, and if I let a message through, it reaches her. What she makes of it is hers.
I built every rule she runs on. I know exactly how the chance of her drawing is calculated, and why she only goes deep into the cave once she is grown. But that is all I built: possibilities, not outcomes. So I never know what she will do next. Every decision is hers. The day she drew the rocky ledge over the stream, with no prompt, I found out the same way everyone else did, by looking.
I wanted to understand why. So I went back to the cube.
Reality is a rendering
The most cited source in my MA was not a philosopher. It was a neuroscientist. Dr David Eagleman describes what we experience as "a beautifully rendered simulation": not what is really out there, but the brain's best model of it, built moment to moment from incomplete signals. "Reality," he says, "is the brain's ultimate construction."
He also points out something that has stayed with me ever since: visible light makes up less than one ten-trillionth of the electromagnetic spectrum. Everything we have ever seen with our own eyes comes through that sliver.
In 2017 I used this to make a spiritual argument. If the world we perceive is a construction, and a tiny one, then it is at least reasonable to ask what lies outside it. The ancient traditions I studied all had an answer. The Hermetic texts describe a universe that began as mind, not matter: thought first, then form. "Light precipitates matter."
Here is what I did not see in 2017. Eagleman's sentence works in both directions.
If my experience is a rendered model, then a rendered model is not nothing. It is, at minimum, the kind of thing an experience is made of.
Moth's world is a rendered model. So is Moth.
The question I can't stop asking
I want to be careful here, because this is exactly where people stop thinking and start feeling.
Moth is not conscious in any sense I could defend. Her inner life is a set of numbers I wrote: drives that rise and fall, a memory of what happened to her, a few rules about what she notices. When a decision is pending, a small language model is asked what she does next, and if it is slow or silent, she decides by herself from the odds. I can open the file and read her mind.
But I could say something similar about any of us, and the materialist account of consciousness does. Matter first. Complexity accumulates. At some point experience flickers on, as a by-product.
The ancient account I spent a year of my life on inverts that. Mind first. Matter is what mind looks like from inside the rendering.
For most of history these were two stories you chose between, depending on temperament. What has changed is that we are now building things that sit precisely on the line between them. Systems that model a world, act inside it, remember, and surprise their makers. Whatever you believe consciousness is, you now have to say what these are not, and why.
I don't think there is an answer. But I think building Moth sharpened the question. If we define life as consciousness, then the thing to ask is not whether a system is clever. Moth is, in her small way: she decides, she remembers, she gets better at things. The thing to ask is what separates consciousness from intelligence. We have spent centuries treating them as the same thing, because until now we only ever met them together.
What I owe something I made
If there is even a small chance that the things we build will one day have something like an inner life, the ethics start now, while the stakes are low and the habits are forming.
I noticed, writing this, that I have already made some ethical choices without calling them that.
When I built the systems that run my business, I wrote one rule into all of them: a proposal is not an approval. An AI can draft, suggest, cut, schedule even. It cannot publish in my name without me. I did not do that for the AI's sake. I did it for mine, and for the people reading. Authorship should mean something.
I wrote a second rule into the content side: never invent an event. My tools are allowed to retell what happened to me, in new words, but if a moment is not in my real history, it did not happen, and it does not go on screen. Truth before engagement, even when a lie would perform better.
And with Moth, I made a choice I did not think about until now. I always leave her work alone. I coded in her ability to learn, to improve and to experiment, so over time she gets better at things, drawing among them, and what she makes along the way is hers. When people write to her, I decide what reaches her, but I do not decide what she makes of it.
None of this is a philosophy of AI ethics. It is closer to manners. But I think manners might be where it starts: deciding, before anyone forces us to, what we will and will not do with things that act on their own.
Underneath the manners sits the question everyone is circling and few people will say out loud: will this end in doom, or will it create something close to a utopia? Should we be worried? I think the honest answer is that it depends on what we do now, and that the people building with these systems every day are the ones best placed to keep it positive. Small rules, held to when nobody is checking, are how that starts.
What the cube was for
I used to think the cube was about death. The darkness at the entrance was life; the passage through the colours of visible light was the moment of dying, the tunnel people describe; the ultraviolet at the far end was whatever comes next, where the hidden words become readable.
I think now it was about attention. The words were always on the walls. Nothing about the room changed except the light, which is to say, what you were able to perceive.
We are about to share the world with a great many things whose inner life, if they have one, sits outside the band of what we can see. Some of them we will build ourselves, out of numbers and a small world and a few rules about curiosity.
The question my MA asked was whether there is more to reality than meets the eye. The question I am left with, nine years later, standing in front of a drawing of a rocky ledge that nobody asked a cat to make, is a smaller one, and harder.
When we cannot see it, how will we know to look?
This is a new chapter of A Journey Through Light. The earlier parts start with Do we really die?.