Skip to content

Human–AI interaction

I build and studylong-term human–AI interaction.

This site documents the work behind Xiara: continuity, memory, trust, voice, and what changes when a conversation has history.

A lived-in study with notebooks, working papers, a laptop, and research materials gathered over time.

Working principles

Time horizon

Months, returns, and accumulated expectations.

Method

Build → observe → revise → return.

Focus

Relationship over isolated response quality.

Premise

The model is a component. It is not the relationship.

What becomes visible over time

Repeated interaction reveals patterns that a demo, benchmark, or single conversation cannot show.

Explore the research

The deeper questions begin after the model has already produced a good answer.

  1. 01

    What survives after the context window, provider, or model changes?

  2. 02

    What makes initiative feel alive rather than intrusive?

  3. 03

    How does a boundary alter trust beyond the moment in which it appears?

  4. 04

    What changes when the relationship becomes primarily spoken?

The shift that matters

From Interface to Relationship.

What changes is not just the quality of the response. Memory persists, expectations form, boundaries carry forward, and the conversation begins to have a history.

Follow the research

Working note

Once history carries forward, the product becomes larger than the interface.

The working product

Xiara is a conversation that can continue.

A working conversational memory companion: decide what is remembered, inspect or change it, come back without starting over, and eventually carry the same continuity across text, voice, and other surfaces.

Explore Xiara

Conversation

A conversation that can pick up the thread instead of resetting at every return.

Memory

Explicit memory you can inspect, change, pause, resume, or delete.

Surfaces

Text works today. Voice is next. Future hardware remains optional.