2026-08-03

Computer-generated algorithmic art is a form of conceptual art. That said, whether a given work is conceptual art depends less on the medium (code, machine, plotter, AI) and more on where the artistic weight is placed.

Since the late 1960s, conceptual art has been defined by one core principle (famously articulated by Sol LeWitt): "The idea becomes a machine that makes the art."

In conceptual art, the idea, system, or instruction is primary. The execution is secondary, even interchangeable. And the work can often be recreated by anyone who follows the rules. Algorithmic art fits this model extremely well.

An algorithm is, in essence, a formalized set of instructions executed by an agent that has no aesthetic intention of its own. This is structurally identical to Sol LeWitt’s wall-drawing instructions. The computer is simply a perfectly obedient, fast, and repeatable draftsman. In that sense, the computer plays the same role as the gallery assistant who draws LeWitt’s wall.

My kind of algorithmic art clearly belongs to conceptual art because the concept or rule system is the artwork. The visual result is a trace or instantiation. Variability, randomness are intentional. The artist accepts loss of direct visual control.

In these cases, the output is not the artwork; it is evidence of the idea.

One difference worth noting is that LeWitt’s instructions are often underspecified, leaving human interpretation whereas algorithms tend toward formal closure.

But this does not disqualify algorithmic art — it simply shifts where ambiguity lives and how authorship is distributed.

On danubedataflow.com I exhibit small pieces based on JavaScript code. Each work has parameters like which color palette is used, the number of horizontal tiles, number of vertical tiles and other parameters that control the algorithm. The user can change these parameters. Each time it is drawn, a new random seed is used. The user can also get a URL that encodes these parameters and the random seed, so they can always get the exact same output.

For inspiration, I sometimes look at early computer art works like those from Vera Molnár, Frieder Nake, Georg Nees etc. I try to find out which algorithm produced this example output and how to parametrize this work.

My art is very clearly situated inside the historical lineage of conceptual, systems-based, and early algorithmic art—much closer to Vera Molnár, Frieder Nake and Georg Nees than to contemporary “generative aesthetics” culture. My self-imposed restrictions are aesthetic axioms typical of conceptual art.

Encoding both parameters and the random seed in the URL effectively creates a certificate of identity and reproducible instructions. It's like a musical score with tempo, key and instrumentation.

This is directly analogous to a Sol LeWitt wall drawing realized at a specific site, or Vera Molnár’s "machine imaginaire".

The fact that users can change parameters but also freeze a specific realization acknowledges both generality and singularity.

My preference for “the whole picture should emerge from a single idea” is philosophically aligned with a mathematical definition rather than procedural storytelling.

Refusing interactivity and animation returns attention to structure, spatial relations and pure description. This echoes early plotter art and conceptual restraint.

For each work, I show a title and a short paragraph describing the rules (not poetic). This echoes conceptual art wall labels. Showing the rules in plain language also reinforces idea-over-execution.

I don't dilute the concept by over-explaining it.

There is no single canonical output. Instead, there are families, parameter spaces and the group that encompasses all possible images defined by a rule.