Algorithmic Archaeology

Van den Dorpel names his method algorithmic archaeology. The first-person definition comes from the Cloud Writings exhibition text (Takuro Someya Contemporary Art, Tokyo, 2026):

"Since 2019, I have dedicated myself to researching artists such as Anni Albers, Vera Molnár, and Charlotte Posenenske, who are known for their systematic approach in creating work with two-dimensional grids, even before the computer as we know it was invented. I develop contemporary algorithms inspired by these historical ones — a methodology I call 'algorithmic archeology'."

Across the generative, digital and NFT fields van den Dorpel encounters a persistent lack of historical awareness, a belief that working with algorithms is something this generation invented. Algorithmic archaeology puts the lineage back in view and, in doing so, separates the conceptual and metaphysical roots of rule-based art from the platform technology that now carries it.

The method runs in two directions.

Excavation. The term first appears in the Spike #70 interview with Tina Rivers Ryan (2022), about the Cartesian Genetic Programming algorithm in Mutant Garden. Rivers Ryan reads the phrase back to him:

"I love this phrase, because it implies that algorithms can be excavated. This means that algorithms are not just pieces of math that are objective and universal and true, but that like fossils, they themselves have a history and are the products of particular historical situations."

An algorithm carries its origin. To use one for art is to find it within its historical layer and redeploy it where its makers did not anticipate. Against the crypto slogan that "code is law," the exchange arrives at a sharper claim: code isn't law but destiny — a legacy that later programmers, and artists, have to build on or around.

Reconstruction. The second direction is forensic. Van den Dorpel researches artists who worked systematically by hand, then rebuilds their procedures as code. The clearest case is Anni Albers, reported in the Angles Morts catalogue (LOHAUS SOMINSKY, 2024):

"While it would have been technically easy to programme an algorithm that replicated her triangle patterns, it turned out that randomly placing the triangles did not achieve the desired aesthetic effect. Van den Dorpel realised that Albers must have formulated clear rules in her head, which she applied algorithmically to each cell of the grid."

The algorithm is deduced from the failure to replicate it without one. Random placement looks wrong, so a non-random rule must have been there, carried as embodied knowledge and never written as code. Making it visible required the random version to fail first, as a diagnostic.

What follows is the lineage as a timeline. It rejects nothing as surpassed. Each algorithm mattered in its moment, and the line runs unbroken to the neural networks and large language models of the present — powerful instruments that are, as algorithms, hard to visualise, because we do not actually understand how they work inside.

Antiquity — the grid, the loom, the mandala

The oldest algorithm is the grid. Weaving is already generative: warp and weft, raised or lowered, one and zero, a rule applied at every intersection until the pattern accumulates into cloth. Jacquard's punched-card loom (1804) is the direct ancestor of the computer punch card. Cuneiform, the oldest writing system, sets wedge-marks in grid-like registers pressed by hand into clay. The mandala and the stained-glass window are rule-based systems in which local decisions accumulate into forms that exceed their individual construction (→ sacred geometry). These are the deep stratum: systematic mark-making as information storage, thousands of years before electricity.

1906 — the Markov chain

Andrey Markov (1856–1922) formalised the chain of conditional probability around 1906: a memoryless stochastic process in which each state depends only on the one immediately before it. It is one of the oldest pieces of usable mathematics in the practice.

Anicca. The Markov chain is the actual mechanism here: each part is chosen conditionally on the part above it, descending through an inheritance graph. A note on the name: "Markov" is also a recurring character in the work — Markov's Window (2004), the Markov's Dream palette, Venster (2024). The namesake honours the mathematician, but not every work carrying his name runs a Markov chain. The chain as technique is most explicit in Anicca, whose conditional inheritance graph descends part by part; Markov's Window uses a memoryless probabilistic rule to permute a seed square. Character and algorithm share a name and are worth keeping apart.

1920s — Anni Albers and the loom as computer

Anni Albers (1899–1994), Bauhaus weaver and textile theorist, understood weaving as structural thought rather than decoration. Her grid is a binary computer, and her patterns follow rules she applied cell by cell in her head. She is the origin case of reconstruction: the random-triangle test failed, so the rule had to be inferred.

Weaving. Albers' embodied algorithm runs forward through the practice: the Anni software works make her cell-by-cell rule explicit in code, scan direction referencing the shuttle's path; Beads carries it into hand-coloured plotter drawings; Stained Unravel hands it to a cellular automaton (→ the grid family: Albers, Anni, Stained Unravel). The Angles Morts catalogue adds a structural argument about who was already doing this: the logic of the grid is the logic of weaving, weaving was for most of its history women's work, and the dominance of women in generative art is continuous with that longer history. "Angles Morts" — blind spots — names what art history failed to see.

1930s–60s — Emma Kunz and the pendulum

Emma Kunz (1892–1963) was a Swiss healer, researcher and artist who made radial line-drawings on graph paper with the aid of a pendulum. She posed a question, read the swings as points and directions, constructed the figure from those readings, then filled the cells with colour, sliver by sliver, a fill that could occupy a full working day per sheet. She understood the readings as coming from a higher order and did not consider the drawings art.

Entfaltung. The excavation runs in stages: re-drawing individual Kunz sheets in code, generalising her construction into a parametrised generator, mapping that generator as a grid, then raising the whole thing into three dimensions and letting the points move over time (→ Entfaltung). Kunz's method matters twice over. It is rule-bound and systematically filled, the same construction logic as the loom and the mandala. And it needs an unpredictable external input, which she located outside herself. That second feature reappears at the end of this timeline.

The drawings on paper. The excavation also runs backward, out of the screen and onto the sheet. The Entfaltung works on paper are drawn with a mechanical plotter and then filled in by hand in pencil, colour laid into the cells one at a time. That hand-fill returns to Kunz directly. She spent up to a full working day filling a single sheet sliver by sliver, and the coloured pencil lying beside this drawing is the same slow instrument. The plotter takes over her radial construction; the hand takes over her fill.

1960s — Charlotte Posenenske and the modular instruction

Charlotte Posenenske (1930–1985) made serialised, modular, reproducible forms that could be assembled and reassembled by others. Her work dissolves individual authorship into structural instruction: the artist specifies a system, not a fixed object. This is the conceptual parallel to the generative protocol, where the artist authors the rules and permits the outcomes rather than composing them (→ "less composed than initiated"). She reached that position without a computer, through Minimal and Conceptual art.

The Interpolations. Van den Dorpel names the Interpolation series (2024) as his Posenenske works. They are plotter drawings on millimeter paper — a pre-printed grid the algorithm draws onto — and they run as a numbered series (I through IX), each a variation of the same construction rather than a singular composition. That is the Posenenske move: a system issued as instances, the individual sheet subordinate to the modular rule that generates it. "Interpolation" names the mathematics of finding intermediate values between known points; the drawings interpolate on the coordinate grid built to measure them.

1940s–1970 — cellular automata

The cellular automaton comes out of John von Neumann and Stanislaw Ulam in the 1940s and reaches its most legible form in Conway's Game of Life (1970): a grid of cells that live, die and reproduce by counting their neighbours. Death is not the system's limit but its motion — cells die by rule, and their death clears space for new configurations.

Stained Unravel. The automaton runs a 23-state morphological vocabulary through positional bitmap compare, temporal colour staining and edge merging (→ 3000, 3001). Senescenence (2026) frames its death not as failure but as the system's natural feature (→ Senescenence), the same reframing that Death Imitates Language applied to genetic organisms. The plotter drawings in that exhibition trace one moment of the automaton onto paper — a negotiation that has since dissolved, fixed in ink.

1968 — Vera Molnár and the first computers

Vera Molnár (1924–2023) is a founding figure of computer art. She worked with rule-based systems before computers were available to her, then with early machines that had no screens and produced only simple outputs she had to translate into physical objects by hand. That translation introduced small imperfections, which van den Dorpel particularly values — not the imperfections of chance but the ones the labour of translation leaves behind.

Shown together. Angles Morts was a two-person exhibition, and the hang stages the argument directly. Van den Dorpel's plotter drawings on graph paper occupy one wall, four small framed works by Molnár the other. The pairing is the point: a contemporary rule-based practice in the same room as the pioneer it descends from, the conversation that should have happened across fifty years taking place on two adjacent walls (→ Angles Morts).

Pression des pairs. The most direct homage in the practice: a generative work built from Molnár's 1959 Quatre éléments distribués au hasard, premiered in the two-person exhibition Angles Morts (2024) alongside her own work (→ Angles Morts). Molnár sits on the hinge of the timeline, the proto-computer figure: systematic before the machine, then constrained by the machine, and appreciated exactly for the human residue the constraint produced.

1970s — Manfred Mohr and the hypercube

Manfred Mohr (b. 1938) has worked since the 1970s with the n-dimensional hypercube: a rigid object with a fixed set of vertices and edges, rotated in four, five or six dimensions and projected to the plane. All the variety in the image comes from rotation and projection while the object never changes.

One object, many figures. Mohr is the nearest precedent in algorithmic art for Entfaltung, which keeps the same invariant — a fixed set of points producing an endless run of figures — but lets the points themselves move, so the object genuinely deforms rather than only turning. This grid of sixteen shows it: every cell is the same construction, the same vertices and edges, re-embedded in space. Where Mohr held to the monochrome line, Entfaltung carries Kunz's colour and sliver-fill onto the moving faces (→ Entfaltung, Stage four).

1974 — the quadtree

The quadtree, introduced by Finkel and Bentley in 1974, is a data structure that recursively subdivides a square into four quadrants, and each quadrant again, as far down as the content demands. It is recursion made spatial.

Struggle for Pleasure. The work is built from a small set of pixel operations — mirroring, rotation, repetition, and in particular subdivision, which is a quadtree: the plane divided into quadrants, recursively. Van den Dorpel's claim about the series is that these plain operations "paradoxically took on an exceptionally complex, transcendental quality not unlike the way the apparently straightforward components of a mandala can," and that "simplicity of operation can become a window into territories of complexity where knowledge itself breaks down" (→ randomness and pattern). The quadtree is where recursion, the grid and the mandala meet in one algorithm.

1975 — the genetic algorithm

John Holland (1929–2015) founded the genetic algorithm at the University of Michigan, set out in Adaptation in Natural and Artificial Systems (1975): a population of candidate solutions bred over generations by mutation, crossover and a fitness function that gives selection a direction. The biological vocabulary is not metaphor imported into computing but the literal operation, abstracted.

Death Imitates Language and Hybrid Vigor. These are the practice's first genetic-algorithm works, and where the fitness function starts to move. In Death Imitates Language (2016) the artist decides which organisms live and die; in Hybrid Vigor (2017) the public decides. Each version encodes a different theory of value, a systematic externalisation of taste into code (→ evolutionary logic). Holland's paradigm is the substrate all of it runs on.

1997 — Cartesian Genetic Programming

Cartesian Genetic Programming extends Holland's genetic algorithm into a graph-based form. It was developed by Julian F. Miller and Peter Thomson at Napier University in the late 1990s for industrial circuit-board optimisation; van den Dorpel credits both Holland and Miller for the lineage he inherits. He encountered CGP in 2019 and wrote his own version in TypeScript.

Mutant Garden. CGP solves the problem that first drew van den Dorpel to it: most mutations of raw code crash, because the halting problem is undecidable, but CGP's graph topology guarantees that no mutation produces an infinite loop, so the system can mutate freely (→ evolutionary logic). Its non-coding genes carry dormant possibilities forward, a genetic archive inside the algorithm. This is the excavation case in its purest form: an algorithm built in the 1990s for circuit boards, found in its historical layer and redeployed for art. "You can see remnants of Windows 3.1; it's still there." Code accumulates; code is destiny.

Contemporary — Tauba Auerbach, systematic by choice

Tauba Auerbach (b. 1981) extends the generative lineage into the present, and does it from the other side of the computer. Auerbach has the machine available and deliberately chooses the manual systematic decision instead — geometric abstraction, optics, weaving, systemic visual logic, worked by hand. Her practice is close to this one in that commitment, though less concerned with producing actual software. She is the contemporary proof that the pre-computer position is a live choice, not a stage the technology left behind.

Now — neural networks and the opaque algorithm

The line runs unbroken to the present, and the present is the neural network and the large language model. These are extraordinary instruments. As algorithms, though, they are hard to visualise, because we do not understand how they function inside: the weights are known, the behaviour is not derivable from them, and no reading of the model explains the output the way a graph of CGP nodes or a Markov transition table does. Every earlier algorithm in this timeline is legible — you can watch it run and read its logic in the image (→ process legibility). The LLM is the first one that is not.

This is where Emma Kunz returns. She said her drawings came from a higher order, using the pendulum as her instrument. Van den Dorpel reads a correspondence between her oracle beyond herself and the large language models he now works with. An LLM is trained on the collective corpus of human knowledge, something close to what Carl Jung called the collective unconscious. It answers from a source larger than the individual, the way her pendulum did. Entfaltung makes the same move structurally, relocating the source again: in place of the pendulum it consults the Ethereum block hash, one ledger of state that every participant agrees on and no one controls (→ the pendulum and the parameter).

Every practice in this timeline needs an unpredictable external input. Kunz called it a higher order. Molnár took it from a constrained machine. The genetic algorithm takes it from mutation, the cellular automaton from its own recursion, Quantizer and Entfaltung from the block hash. The LLM takes it from the collective corpus. The dissociations wiki itself is the current instance: this page, and the archive around it, is maintained together with an LLM (→ about). The oracle changes address across a century. The mechanism — getting the deciding mind out of the loop so a structure the maker could not have authored can appear — does not.