Wide but Shallow: Coding Agents in Generative Art
This paper presents an autoethnographic study on the use of coding agents in code-based generative art. Over the course of the 31-day Genuary art challenge, the author examined and compared a manual approach with the use of coding agents in their practice. The growth in quality of coding agents enables their use to autonomously produce programming code. Generative art's bricolage quality of combining existing algorithms and a coding agent's ability to draw on extended training data leads to a difficulty collapse: nearly any well-known algorithm can be used directly. While this enables easier free-form experimentation and a wider variety of outputs, it comes at the cost of less depth. Overall, this points towards a slow progression in artists' roles from inventing algorithms to changing parameters of existing ones and a potential overall artistic surrender to the results from large-language models.