A generative work can stay active for months, produce thousands of configurations and reach the end without retaining anything of what happened. We have learned to ask how these systems produce the present, but not what relationship they build with their own past. Imagine a generative work in a museum. It keeps running for weeks, perhaps months. Every day it produces images, configurations, variations that the public watches for a few minutes. The system goes on. In the end, the exhibition closes. What remains in the work of all that time? Often nothing. The last state produced does not necessarily carry any consequence of the first. The months that went by belong to the history of the exhibition, not to the structure of the work. The machine had a duration, but it did not become experience. It ran for a long time without building itself a biography.
We are so used to this condition that we consider it almost natural. But it is not. A generative system that forgets what it produces was built to do so. It is a design choice. And perhaps we should begin to weigh its consequences.
Residual Time and generative art: the past we do not ask about
In recent years the critical vocabulary around generative art has grown rapidly. We have learned to ask many questions, necessary ones, to distinguish practices grounded in the construction of systems from the mere production of technologically spectacular images.
And yet almost all of them consider the work by fixing it at a given moment. We rarely ask what happens to a form once it has been produced. Does it disappear? Is it archived? Does it go on having some effect? A work can be sophisticated, rigorous, and still live in an absolute present. Each new configuration is born, occupies the field of vision for a few moments, disappears and gives way to the next. The past is consumed almost as fast as it is produced.
Amnesia as a condition of generative art
I do not necessarily consider this amnesia a limitation. Algorithmic spectacle lives on transformation. It must be able to recombine its visual space quickly. The system can keep doing so precisely because the weight of what has already happened remains small.
If every event produced a significant consequence, after weeks or months the work would find itself bound by its own history and would have less freedom. Memory introduces friction. It is therefore possible that amnesia is not a defect but one of the conditions that make the system efficient. The problem begins when this condition is taken for granted even in works that claim to treat duration as their material. Lasting and accumulating are not the same thing.
A time that passes without settling
The live simulations of Ian Cheng have taken the idea of the work as process very far. In the Emissaries trilogy the viewer faces a world already under way, which will go on after the visit. It is an important step: the time of the work stops coinciding with the time of the viewer’s individual experience.
But a simulation can continue indefinitely without truly accumulating its own duration. Time passes, but it does not settle into the system.
Long-form generative art has approached the problem in an almost opposite way. The process takes place, generates a result, and that result is fixed. Becoming produces a stable object. It is an effective solution as far as the circulation of the work is concerned: the result can be identified, certified, collected. But time is resolved by stopping it. On one side we have a present able to continue indefinitely. On the other, an instant taken out of the process and made permanent. Two very different possibilities, with a shared question: at what moment does what has already happened truly begin to change what can happen next? It is from this question that, for me, the concept of Residual Time arises.
Residual Time in generative art
Residual Time does not coincide with duration. A system can run for years without producing any. Nor does it coincide with memory in the computing sense of the term. An archive can preserve every single previous state without any of those states still having the slightest effect on the machine’s future behavior. I speak of Residual Time when what has happened continues to act beyond its original event. A form appears and disappears. Something, however, remains as an operative trace that can modify a probability, alter a threshold. The past then stops being merely what precedes the present. It begins to work inside the present.
In this sense the residue is what keeps acting. We can recognize at least three conditions. The first is persistence: something that survives the event. The second is resonance: a trace that enters into relation with other states, other traces or new material, acquiring a meaning that also depends on relations built later on. The third is activation: what has remained latent must be able to return and modify the behavior of the system or the form it produces.
The combination of these three conditions introduces a causal continuity. The new state no longer arises exclusively from the initial rules, but also from what the system has already been. The future stops starting from zero. Every configuration carries within it a deformation produced by the previous configurations. This is where duration becomes history.
Remembering also means choosing
A machine capable of preserving everything does not necessarily have a meaningful relationship with its own past. Indiscriminate memory is an archive. For the residue to take on operative value, a selection is needed, one that may depend on design decisions. Someone has established the conditions that allow a trace to survive.
I call this step selective reinjection. It is the moment when what the system has produced can return into the process and modify it: this is where memory meets authorship. The artist can decide which part of the past will keep the power to influence the future. A protocol can iterate indefinitely while remaining identical to itself. A residual system changes also because something has already happened to it.
When the output returns into the system
This condition produces a further theoretical consequence. In Languages of Art, Nelson Goodman distinguishes between autographic and allographic arts. In the former, the identity of the work is tied to the specificity of the object. A perfect copy of a painting does not thereby become the original. In the latter, there is a notation capable of producing many correct performances. Music is the most obvious example.
Much of conceptual art and of instruction-based artistic practice can be read through this structure. In Sol LeWitt’s wall drawings the work can be made again because the instructions exist. The model works as long as the score remains stable.
But what happens if the performance modifies the score? In a residual system the output can re-enter the generative conditions. What is produced today modifies what the system will be able to produce tomorrow. The separation between rule and result becomes less clear-cut. Performance is no longer only a consequence of the protocol; it can gradually become one of its components. At each pass, the system inherits something from the previous configurations. At that point we have a notation that is rewritten by its own history. The allographic category begins to prove insufficient. The work does not coincide with a single material object, but neither does it coincide with an immutable score. Its identity shifts into the relationship between rule, memory and transformation.
When one residue meets another
So far we can imagine all this within a single system; but the residue can enter into relation with material coming from outside. At that point the nature of the problem changes. We no longer have only a machine transformed by its own past. We have different processes that begin to modify one another through what they leave behind. A trace produced at one point can become a condition for something that will happen elsewhere. What emerges then carries within it histories that do not entirely belong to it.
Time, from a succession of events, becomes a field of relations. This shift also changes the meaning of the word “residue.” The residue can become an agent of transformation. What remains of one process comes into contact with what remains of another, and neither continues exactly in its previous form. The relation produces new conditions. At that point we have to ask what can arise from the interaction between different memories.
Residual Time as a critical problem of generative art
In front of a generative work we should ask: what does this system remember? And immediately after: what has it decided to forget? The problem becomes evident when material from outside enters the system. Imagine processes fed by geopolitical events, environmental crises, economic fluctuations, conflicts, migrations, social transformations. Memory immediately stops being neutral. For how long should a war continue to weigh? Must an event from yesterday necessarily carry more force than one that happened six months ago?
We can describe all these decisions through thresholds, weights, coefficients and decay curves. The language remains technical, but the act is not. We are establishing which parts of reality will continue to have consequences. We are building a hierarchy of memory.
This is where the problem of Residual Time also becomes a critical problem, one that calls for reflection on the conditions under which something continues to carry weight while something else is allowed to disappear.
But there is a further step. When traces coming from different processes begin to interfere, memory becomes a shared condition. What one system leaves behind modifies the environment in which another system will have to act. Perhaps Residual Time truly begins here: when what survives becomes part of the conditions through which something else can come into being.
Dario Buratti