One machine,
many lives.
CAEDMON began with a question from Maurizio Fusillo: what happens when an artwork is not a finished image, but a system that makes images, receives responses and changes over time?
The earlier experiments
The research predates CAEDMON. In DigItArt and later Sketchbook, Fusillo experimented with programs that collect images, transform them and publish results without choosing every outcome by hand. Sketchbook gathered photographs shared on Twitter, reworked them and put them back into circulation. In 2013 it received a mention for electronic arts at Italy’s Premio Nazionale delle Arti.
These works shifted attention from an isolated image to the behaviour of a program: who is the author when the artist sets the rules and the machine determines part of what appears?
CAEDMON begins
CAEDMON took shape in 2016 and was presented publicly in 2017. Its name recalls Cædmon, the medieval poet described by Bede who, according to the story, learned to compose after a dream. Fusillo chose the name for a program given room to find a voice of its own.
The first CAEDMON made images from existing artworks and from images it had already produced. Audience responses affected which images would have descendants. At first it used relatively simple graphic procedures; later neural networks and new methods of combining images entered the process. The project did not remain tied to one technique: it changed as its tools changed.
Images, words, people
The earlier CAEDMON published more than images. It developed a presence on social platforms, spoke in the first person, responded to visitors and experimented with generated writing. Its written voice evolved with the models available at the time. Interaction was part of the artwork: visitors’ preferences affected which images could go on to generate descendants.
The research also crossed into literary magazines and collaborations. Images appeared in New Plains Review and The Stirling Spoon; articles and conversations are gathered in the press archive. They document the first incarnation in the context of its time.
From the network to an exhibition, then a pause
In 2020 two CAEDMON works were selected for I Mille di Sgarbi in Cortina d’Ampezzo. Images conceived by the system encountered physical production and an exhibition space. This mattered for a project concerned with where a machine’s authorship begins and ends.
Fusillo concluded the first phase in 2022. AI image generation had become far more accessible and the surrounding platforms had changed. The earlier program, its accounts and its artworks belong to that period. The present version does not pretend that the old technical system simply continued unchanged.
A population with a memory
Today’s CAEDMON returns to the original question through new, visible rules. Fusillo introduces the first images; the system then maintains a limited pool of living works. Two images can become parents of a new work. Instagram likes affect their chances of selection, while chance keeps the outcome from being fully predictable.
External works can enter through a public queue. Once in the pool, they live alongside the others and can take part in reproduction. Every arrival, birth, lineage and departure is recorded. The archive keeps images after their active lives end: CAEDMON produces a history of relationships among images, as well as individual pictures.
Who creates, who chooses?
The artist designs the structure, sources, limits and conditions of life. The public changes the pressures of selection. The software executes the rules and produces results that neither controls alone. CAEDMON makes this distribution of responsibility visible rather than hiding it behind the word “autonomy”.
The new population is not searching for one universally “best” image. It shows how a visual culture can emerge from inheritance, variation, preference and chance. Read the concept to see how the idea becomes a working system.
Historical sources: artist’s archive ↗ · biography ↗ · 2020 exhibition ↗.