Part of Temporal Constructs Appendix, which contains details behind the research: dates, quotations, descriptions of instruments, and sources.

General definition

An agent is anything that perceives its environment and acts upon it (Russell and Norvig). An AI agent today is usually a software system built around a language model. It pursues goals by planning and taking actions with tools, and it holds partial autonomy from step-by-step human instruction.

Relation to the practice

Agents act in the symbolic domain only. They read manuals and documentation, program legacy languages, author and recall presets, and operate instruments through a shared command language, without hearing the result. Each agent works on a specific instrument or task, coordinated by an agent that keeps the context of the whole system. The artist listens and judges.

The agents are both collaborators in the work and objects of study, and some are self-evolving frameworks that develop their own skills over time.

Dating of claims

Every claim in the account concerns the models and arrangements the practice worked with, with the date attached. The practice ran a changing set of models of different sizes through 2026, and some of them ran locally. What the practice learned from one arrangement did not carry over to the next. A model that failed at one task was the better choice at another, and the models on offer changed without notice.

Unforeseen

Here the unforeseen situations, sometimes difficult, sometimes unexpectedly interesting, will be described as the project runs. Many of these unforeseen situations have already occurred and have been recorded. They will be added in the time to come.