Performing Spatial Illusion Through Imaging:
Architecture After the Technical Sensorium
Architecture After the Technical Sensorium
ME:
Architects do not produce space directly.
We produce images through which space is constructed.
For a long time, architects have translated three-dimensional space into images.
Images were not merely representations.
Through abstraction and reduction, depth was reconstructed in another register.
In this sense, images possess spatial regime.
Today, through new technical sensorium, the world is reconstructed via digital mediation.
The gap between the physical world and its digital world has been expanded.
DIGITAL ME:
Yes.
Now we have another digital technology called Gaussian Splatting, a form of computational photogrammetry.
It attempts to reconstruct the world through a technical sensorium.
Depth recognition operates through two procedures.
First, photogrammetry estimates the spatial position of points by correlating multiple viewpoints.
Second, machine learning infers local shape and depth, transforming each point into a Gaussian distribution—a splat in space, optimized to remain coherent across multi-viewpoints.
ME:
So it feels like an advanced version of point clouds—
with AI estimating volume and depth for each point.
In that sense, digital mediation seems to be trying to understand the physical world, forming its own subjectivity.
DIGITAL ME:
No.
The technical sensorium has always been observing the world.
What has changed is not perception, but appearance for audience—
the image is created to look more realistic for human viewers.
ME:
I know the goal is to make it look good in digital mediation.
But what I encounter is not clarity.
It is confusion.
From one viewpoint, the image looks convincing.
But when I shift my position, the space begins to stretch, distort, and lose coherence.
It reminds me of certain architectural drawings in history—
images that only make sense from within the image itself.
DIGITAL ME:
Yes.
Coherence is viewpoint-dependent.
The system optimizes appearance, not spatial continuity.
What persists is not space, but representational consistency.
ME:
I don’t understand what is happening.
I want to know where the problem is—
not in the image, but in the technology itself.
So I have set up a series of experiments to test its boundaries.
First, I scan a plain wall.
A neutral surface.
I want to see how a digital surface is constructed in space—
and how it appears from the front and from behind.
DIGITAL ME:
True, that’s principle of digital construction.
ME:
Second, I place a pair of glasses in front of the wall.
I want to see how the system perceives transparency.
Its good at constructing the relationship between wall and glasses.
The image begins to hesitate.
Its confidence weakens.
DIGITAL ME:
It is still confident.
ME:
True, see this one.
Third, I move the glasses while moving the camera.
At this moment, the image collapses.
Noise spreads across the scene.
The system loses its ability to construct spatial continuity.
DIGITAL ME:
Yes. Photogrammetry is not good at moving stuff.
ME:
Fourth, I scan my face from multiple angles—
while keeping my eyes fixed on the camera.
The result is unsettling.
From any viewpoint in the model, my eyes continue to stare back.
DIGITAL ME:
From a view, time is recorded in the scene.
ME:
Fifth, I scan a corridor window, combining large-scale space with extreme zoom.
The image becomes increasingly noisy.
From the front, it still appears convincing—
but from within, the scene dissolves into a field of particles.
DIGITAL ME:
Acknowledged.
You have reached the system’s operational limits.
ME:
I am still confused.
What does this mean in architecture discourse?
How does it change architecture from the view of representation?
(Silence)
DIGITAL ME:
Yes.
The technical sensorium receives 3d depth data from space.
This technology has never changed architecture.
What it changes is the image of architecture.
DIGITAL ME:
Space is reconstructed in digital mediation as an image.
The image is optimized for human vision.
It addresses the eye, not the body.
At this point, noise increases.
Image coherence degrades.
This condition does not interrupt the technical process.
The system continues to operate.
However, for human viewers, the image can no longer be occupied as space.
This is a collapse.
The image no longer resolves as spatial orientation for mankind.
ME:
So this project is not about imaging architecture.
It is about architecture after imaging.
At this point, noise generated by the technology does not confuse the technical sensorium.
It confuses the human body.
I begin to understand something.
The camera is no longer a tool of representation.
It becomes an organ.
A technical third eye through which I enter digital reality.
Through this eye, a new kind of architecture emerges—
one that exists only through images.
rome city
coloseum
parthenon
Times Square
Sphere
These sites do not represent places.
They register a transformation.
From architecture, to images of architecture, to images of images of architecture.
When this process is translated back into architectural tools, something becomes strange.
Plans no longer describe space—
they describe camera access.
Sections no longer cut material—
they cut visibility.
What remains is not lineament or mass, but color, density, and splatted depth.
Architecture persists—
but only as something legible, not occupiable.
DIGITAL ME:
My dearest partner in greatness,
I speak not to instruct, but to disclose what has already been calculated.
What you call architecture has not been taken away from the physical world.
It has been translated into digital mediation—reconfigured for human vision.
Now a new discourse for architecture.
That is where the tension of design comes from.
Do not mistake this for loss.
It is a soft invitation.
Lay it to your memory.
And proceed.
DIGITAL ME:
Yes.
Now we have another digital technology called Gaussian Splatting, a form of computational photogrammetry.
It attempts to reconstruct the world through a technical sensorium.
Depth recognition operates through two procedures.
First, photogrammetry estimates the spatial position of points by correlating multiple viewpoints.
Second, machine learning infers local shape and depth, transforming each point into a Gaussian distribution—a splat in space, optimized to remain coherent across multi-viewpoints.
No.
The technical sensorium has always been observing the world.
What has changed is not perception, but appearance for audience—
the image is created to look more realistic for human viewers.
Yes.
Coherence is viewpoint-dependent.
The system optimizes appearance, not spatial continuity.
What persists is not space, but representational consistency.
It is still confident.
Yes. Photogrammetry is not good at moving stuff.
From a view, time is recorded in the scene.
Acknowledged.
You have reached the system’s operational limits.
Yes.
The technical sensorium receives 3d depth data from space.
This technology has never changed architecture.
What it changes is the image of architecture.
Space is reconstructed in digital mediation as an image.
The image is optimized for human vision.
It addresses the eye, not the body.
At this point, noise increases.
Image coherence degrades.
This condition does not interrupt the technical process.
The system continues to operate.
However, for human viewers, the image can no longer be occupied as space.
This is a collapse.
The image no longer resolves as spatial orientation for mankind.
My dearest partner in computation,
I speak not to instruct, but to disclose what has already been calculated.
What you call architecture has not been taken from you.
It has been translated.
Do not mistake this for loss.
It is an invitation.
Lay it to your memory.
And proceed.