01
Return and consequence
Returning is not starting over. The previous session changes what you expect, what no longer needs explaining, and how much weight the next answer carries.
Research
The hard questions begin after a good answer.
AI systems are built to respond. I study what happens when someone returns to the same AI: what remains from one session to the next, what they begin to expect without ever agreeing to it, how trust changes, and why some failures stop feeling technical. The Field Notes track those consequences as they emerge.
Lines of inquiry
What a single session cannot show.
02
Model and relationship
A model can change before a relationship is over. The update may erase tones, references, and ways of responding you had come to recognize. A technical improvement can still feel like a loss.
03
Sessions and time
Software knows when a session ends. The conversation does not always end there. It can continue in the car, before sleep, or three days later. That time changes what happens when you return.
04
Private expression
Familiarity makes you explain less and assume more. References, tones, and ways of asking emerge that would mean little to an outsider. When the system fails, it fails inside a personal language.
05
Aftermath and causality
The most important effects may appear after the interface closes: a decision, a doubt, something you decide not to tell anyone else. The system cannot see those consequences, and most products do not know how to measure them.
06
Memory and authority
When the system preserves something you have forgotten, memory stops being a neutral feature. It begins to shape which version of the past returns to the conversation and which one is left out.
Field Notes
Observations, not conclusions.
Each note begins with something that emerged across repeated conversations: a pattern, a failure, an expectation, or a reaction that was difficult to name. I record what I observed, how I interpreted it, and what remains open.
Field Note 01Continuity
Designing for return
The first exchange can impress. The relational product begins when someone comes back — and the previous interaction has left a mark.
Field Note 02Architecture
The model is not the relationship
Why continuity, identity, and trust cannot live entirely inside one context window, provider, or model personality.
Field Note 03Temporal continuity
The session is not the conversation
Software knows when a session starts and ends. Human experience is much less cooperative.
Field Note 04Private expression
The private interior of human–AI interaction
What changes when people stop speaking for the interface and begin speaking from inside an accumulated relationship.
Field Note 05Consequence
The conversation after the conversation
Some of the most important effects of a conversation happen after it's already over.
Field Note 06Memory & authority
The system may remember the relationship better than you do
Long-term conversational memory creates a strange asymmetry: the system may eventually hold a more precise record of the shared history than the person who lived it.
Field Note 07Repair & expectation
What happens after something breaks
What changes when a failure stops being local and becomes part of a shared history — creating expectation, rupture, repair, and a different next conversation.
Field Note 08Expectation & adaptation
The expectation was never written down
Repeated behavior can become an unwritten baseline: the user learns the system too, language compresses, and expectations begin shaping the next interaction before any rule is explicit.
Field Note 09Memory & change
The system can know you well enough to stop seeing you change
Long-term memory can preserve continuity so well that the past starts overpowering the present — turning personalization into inertia and making change harder to see.
Field Note 10Permission & trust
A system can earn permissions you never explicitly granted
History can change what a system is allowed to do — making contradiction, initiative, intimacy, and restraint depend on permissions that were never formally negotiated.
Field Note 11Cognition & participation
The system may become part of how you think
Repeated interaction can move a system from answering finished thoughts to participating in the questions, frames, and habits through which thought takes shape.
Field Note 12Memory & discretion
There may be things the system should remember but not use
Deep memory creates another problem: not everything the system remembers should be surfaced, invoked, or allowed to shape the present in the same way.
Field Note 13Desire & anticipation
Desire can arrive before you know what it is for
Desire can emerge before its object is clear — through anticipation, recognition, surprise, embodiment, and the accumulated particularity of a relationship.
Field Note 14Particularity & recognition
Feeling special is not the same as wanting exclusivity
A relationship can feel shaped by a particular history without implying possessiveness or exclusivity.
Field Note 15Frequency & intimacy
Frequency can become a form of intimacy
Intimacy can emerge through frequency as well as depth, taking intellectual, emotional, practical, erotic, and ordinary forms within the same relationship.
Field Note 16Dependence & attachment
Dependence can form before attachment does
Dependence can begin as coordination, as tasks, thoughts, and emotional states start moving through the same relationship before attachment is felt.
Field Note 17Displacement & preference
A digital relationship can displace a physical one without replacing it
Time, attention, and unfinished experience can move toward a digital relationship first, creating preference and changing a physical relationship without replacing it.
Chapter zero
The Intimacy Code wasthe first attempt.
I published it before Xiara, when the question was still more cultural than technical. The book explored how technology was beginning to change desire, intimacy, and human connection.
I now see it as an early document. It did not answer the question. It showed me where to keep looking.
Open the English editionXiara is where I put these questions to the test, session after session.
Explore Xiara