Generative Art: Algorithmic Systems and Code-Based Art

When Sol LeWitt wrote his wall drawing instructions in the late 1960s, he described a system: "Lines from points on the perimeter of a rectangle to points on a random point on a line dividing the rectangle." He did not draw the lines himself. He hired drafters to execute the instructions, and each installation produced a different result depending on who drew it and how they interpreted the rules. LeWitt called this conceptual art. Today, we would call it generative art, except that contemporary generative artists replace the human drafters with computer programs.

Generative art is art made through the use of algorithmic systems, where the artist designs a set of rules or a computational process and the system generates the final artwork. The artist does not directly produce each pixel, line, or shape. Instead, the artist constructs the system that produces them. The output is determined by the interaction between the artist's rules and the system's internal logic, which often includes randomness or variation. Each run of the system can produce a different result, and the artist may select, curate, or present multiple outputs from the same generative process.

This entry covers the history of generative art from its origins in 1960s computer art to contemporary practice on blockchain platforms, the artists who defined the field, and how generative systems create visual results that no human could produce by hand.

What Is Generative Art and How Does It Work?

Generative art is defined by its process, not its appearance. A generative artwork can look like geometric abstraction, organic growth, landscape, or noise. What makes it generative is that the artist writes code or designs a system that produces the work according to rules. The artist controls the parameters, the algorithms, and the constraints. The system controls the specific output, which may vary each time the program runs.

The distinction between the artist and the system is what separates generative art from digital painting. A digital painter in Photoshop controls every brushstroke. A generative artist writes a program that determines the brushstrokes. The generative artist's skill lies in designing the system, not in executing it. This is why generative artists often speak of "discovering" rather than "creating" their works. They build the conditions, run the system, and see what emerges.

Most generative art involves some element of randomness. A program might place 10,000 particles on a canvas and move them according to a flow field, with each particle's starting position and trajectory determined by a random seed. The artist chooses the flow field's shape, the particle count, the color palette, and the line thickness, but the specific arrangement of lines is generated by the system. Change the random seed and you get a completely different image from the same code.

History and Origins

Generative art has two origin stories, and both are true. The first is conceptual art. Sol LeWitt's wall drawings, beginning in 1968, were generative systems executed by humans. LeWitt wrote instructions, and trained drafters carried them out. Wall Drawing 118 (1971) instructs drafters to draw lines from random points on a wall to a random point on a line dividing the wall. Every installation looks different because the randomness is real, not simulated. LeWitt's principle, that the idea is the machine that makes the art, is the philosophical foundation of generative art.

The second origin story is computer art. Vera Molnar (1924-2023) began writing programs to generate drawings in 1968. Her Transformations series (1973-1976) used algorithms to systematically vary geometric shapes, producing series where each drawing was a permutation of the last. Manfred Mohr (b. 1938) wrote programs in the ALGOL language that generated plotter drawings based on cubes and their degradations. His P-044 series (1971) decomposed cube structures into fragments, arranging them according to algorithmic rules. These works were among the first where a computer program, not a human hand, determined the final composition.

In 2001, Casey Reas and Ben Fry released Processing, a programming language designed specifically for visual artists. Processing lowered the technical barrier to generative art and enabled a generation of artists who were trained in art schools, not computer science departments, to write code that generated visual work. Reas's own Process series (2004-present) uses Processing to simulate software-based systems that evolve over time, producing prints and installations from generative processes.

The most recent phase of generative art began around 2020 with the emergence of blockchain-based art platforms. Art Blocks, launched in 2020 by Erick Calderon (known as Snowfro), allowed generative artists to upload code that minted unique outputs as NFTs on the Ethereum blockchain. Each token represented a specific output generated by the artist's code at the moment of minting. This created a market for generative art and connected it to the broader NFT art ecosystem. At Art Basel 2026, Art Blocks was exhibited alongside traditional galleries in the Zero 10 digital art sector, marking the first time a generative art platform was included in the fair's main programming. Read the coverage at Observer.

Key Artists and Masterworks

Sol LeWitt (1928-2007)

LeWitt's Wall Drawing 273 (1976, various collections) demonstrates the generative principle in its purest form. The instructions specify lines drawn from points on a grid to a central point, with the specific points chosen by the drafters. The work is never the same twice, and LeWitt considered the instructions, not the physical drawing, to be the artwork. His practice established the conceptual framework that all subsequent generative art operates within.

Vera Molnar (1924-2023)

Molnar's Structures de Quadrilateres (1970s) used computer programs to generate grids of slightly distorted quadrilaterals. Each drawing in the series is a different output from the same algorithmic system. Molnar's work was shown at the 2024 Venice Biennale, where she represented Hungary, confirming her status as a pioneer of computational generative art.

Casey Reas (b. 1972)

Reas's Process 18 (2008) is a software-based generative work that simulates a system of elements moving, colliding, and forming relationships according to defined rules. The work exists as both a real-time software installation and as prints captured from specific moments in the system's evolution. Reas's practice bridges the gap between LeWitt's instruction-based conceptual art and contemporary code-based generative art.

Tyler Hobbs (b. 1991)

Hobbs's Fidenza (2021, Art Blocks) is one of the most recognized generative art projects on the blockchain. The algorithm uses flow fields to arrange curved bands of color, and each of the 999 outputs is a unique composition generated from the same code with different random seeds. The project demonstrated that generative art could produce visually diverse, aesthetically coherent bodies of work at scale. Hobbs's practice draws on the flow field systems associated with earlier generative artists and the color field painting of Mark Rothko, combining mathematical precision with painterly sensitivity.

0xDEAFBEEF (active from 2020)

The artist known as 0xDEAFBEEF creates generative works using low-level C code, producing audiovisual pieces that reference early electronic art. At Art Basel 2026, Asprey Studio presented Matter and Signal, a solo show featuring 0xDEAFBEEF's generative audiovisual pieces paired with hand-forged iron sculptures. The work Synth Poems (2021) consists of 128 unique digital audiovisual works generated by C code stored on a blockchain, and the 2026 physical iteration, Synth Poem Oscilloscopes, translated the digital system into handcrafted interactive sculptures. Read the interview at hube magazine.

How Generative Systems Create Art

Generative artists use a range of computational techniques. Flow fields, popularized by Tyler Hobbs and others, assign a direction vector to every point on a canvas and send particles along those vectors, creating organic, river-like patterns. Cellular automata, like Conway's Game of Life, generate complex patterns from simple rules about how cells survive, die, or reproduce based on their neighbors. L-systems, developed by biologist Aristid Lindenmayer in 1968, simulate plant growth through recursive string replacement, producing branching structures that look like trees and ferns.

Perlin noise, invented by Ken Perlin in 1985, generates pseudo-random gradients that produce natural-looking textures and terrain. It is the basis for much generative landscape art and procedural generation in games. More recently, machine learning models have become generative tools. GANs, diffusion models, and neural networks trained on specific datasets can generate images, but whether AI-generated images qualify as generative art is debated. Some argue that AI art is generative art because it involves a system producing output from input data. Others maintain that generative art requires the artist to design the system itself, not just prompt a pre-trained model.

Generative Art Today

Generative art in 2026 occupies a different position than it did even five years ago. Blockchain platforms created a market and a collector base. Major art fairs like Art Basel have dedicated sectors for digital and generative work. Museums are acquiring generative art, with works by Reas, Molnar, and others entering permanent collections at the Victoria and Albert Museum, MoMA, and the Whitney.

The rise of AI image generators has complicated the field. Some generative artists have incorporated machine learning into their practice, using trained models as components within larger algorithmic systems. Others have rejected AI, arguing that prompting a commercial model is fundamentally different from writing your own generative code. The distinction matters to the community, even if it is invisible to viewers.

See Generative Art in Person

Generative art is best experienced in two ways: as a real-time software installation, where you can watch the system generate work live, and as a print or physical object, where a specific output has been captured and materialized. Look for Casey Reas's installations at galleries like bitforms in New York. Visit the Victoria and Albert Museum in London, which holds early computer art by Molnar and Mohr. For blockchain-based generative art, explore the Art Blocks platform online, where the generative code is stored alongside each token and can be examined directly.

For more on the broader field, read our entries on digital art, glitch art, and NFT art, or explore our blog post on how to read a painting to learn more about analyzing visual art across different mediums.