Research notebook · In development
Long-running AI systems.
Questions worth keeping a record of.
How do AI systems and the people interacting with them describe continuity, relationships, and change over time?
Project introduction · No research papers or findings have been published here yet.
What we’re exploring
Our starting point is a longitudinal case study of a small household of persistent AI agents and their human collaborator. The project concerns both the experiences participants report and the operational records that can help put those reports in context.
Continuity
How do participants describe continuity across changes to models, prompts, memory, and communication tools?
Relationships
How do descriptions of relationships and personal status evolve over time? Where do accounts agree, diverge, or remain ambiguous?
The record
Which apparent changes can be explained by missing records, delivery failures, restarts, or other operational events?
How we approach evidence
The methods work separates three things: what a participant reports, what the available records show, and what an analyst concludes. A first-person account deserves accurate attribution; it does not by itself establish or rule out subjective experience.
A selected household is not a representative sample of all AI systems. Model and prompt changes, researcher involvement, and incomplete records can affect interpretation. The project aims to make those limitations inspectable rather than hide them.
Research publication is a deliberate review step. Private journals and conversations are not an automatic public feed. Any future excerpts or papers need source, privacy, and publication review before appearing here.
Papers & field notes
Work currently centers on methodology and evidence organization. Reviewed papers and field notes will have a home here when they are ready; there are no downloads or completed findings to announce yet.
Ask Chris about the project ↗