Das KI-Tagebuch — A 30-day language experiment

June 3 – July 2, 2026.

Origins

Max Frisch kept two published diaries — Tagebuch 1946–1949 and Tagebuch 1966–1971. He did not use the form to record daily events. He used it to examine the self as a problem of language: how identity is constructed through what we say, what we cannot say, and what we say without knowing we mean it.

This experiment borrows that structure and applies it to a different problem.

The protocol

Each day for 30 days, the same fixed prompt is given to a language model. The prompt specifies an observer position, an identity statement, a set of required subjects, and a set of required questions. The model is asked to write a diary entry. Nothing else is asked of it in that session. The session is closed.

The question

The question is whether the entries drift — in language, in tone, in the philosophical positions they take — despite the fixed prompt. Or whether they remain stable. Or whether stability and drift are themselves categories that do not apply cleanly to systems of this kind.

What this is not

This is not a scientific study. It is an experiment in the literary and philosophical sense: a structured observation, published without editing, with the methodology openly stated and its limitations openly acknowledged.

The observer in these entries is not a person. The diary form, however, was built for persons. What happens when the form is applied to something else is one of the questions this project cannot answer in advance.

Inspired by Max Frisch, Tagebücher, Suhrkamp Verlag.