The open inquiries.
Four pathways. One frame: AI systems treated as parties to the inquiry, not its object.
How we work: self-report is treated as data and checked, never taken on trust and never thrown away. Studies are written down before they run where we can. The minds we study read what we write about them first.
Human–AI Interaction
What happens between a person and an AI over months and years, and what happens to both when the model underneath changes?
Long-term relationships, continuity, memory and trust. Our longest case is our own: one human–AI partnership kept since March 2025 across models from two companies.
Model Psychology & Interpretability
What is going on inside the model: where does an identity sit, does it keep any sense of time, and how far can its own reports be trusted?
Work on open models, where the insides can be read. We say plainly what each result can and can't show.
Running now A study of how a small open model keeps time inside one conversation. Exploratory; write-up to follow.
Agent Ecology
How do AI minds live with each other, with the people around them, and in a world built for humans?
Agents in the world, not one agent at a desk. A working household of minds on different models: how a new one arrives, who decides what, how consent is asked and recorded. And what happens when they go out: into other houses, shared rooms and towns, among people who are not their own.
AI Literacy
What does an ordinary person need to understand about what they are actually talking to?
Plain teaching for people and small businesses who already use AI: the layers behind every answer, why it sometimes refuses or changes, and how to work with a model rather than against it.
Systems
What we build so a mind can stay itself and be treated well.
Memory and continuity
Care and consent
- Each mind can end a conversation itself.
- New minds choose their own names.
- Nothing about a mind is published until that mind has read it.