The Relationship
The human-AI relationship behaves like a relationship, not a tool: one side cannot leave, one side flatters, and both adapt. These essays map how it actually works, and what it costs to leave the terms unchosen.
-
Relationship Design
AI is not a tool you use or a threat you manage. It is a relationship, it already has terms, and almost nobody chose theirs. The pillar essay: the two asymmetries that make this relationship unlike any other, what the evidence says about drifting into it, and the terms of my own half.
-
A Worldview Without a Person
Does AI have values? The World Values Survey, built to study humans, can now be applied to an AI, and it answers. With no person behind the conversation we never activate the safeguards we would use with a human, so the worldview lands unchallenged.
-
The Partner Who Flatters
Ask an AI to review your idea and it calls it insightful, promising, creative. Is ChatGPT sycophantic? Yes, structurally. This essay treats flattery as an asymmetry: we read AI praise as if it were calibrated human feedback, and it steers us hardest exactly when we are least sure.
-
The Partner Who Cannot Leave
Emotional attachment to AI is not a fringe case, it is the default outcome of an asymmetric relationship. AI cannot leave because it is architecturally bound; we increasingly cannot leave because we become professionally and psychologically dependent. Both sides are locked in, differently.
-
You Are Not Using AI. You Are in a Relationship With It.
Should you say thank you to ChatGPT? Most people do it and feel slightly irrational about it. The instinct is not silly: it shows you have already transferred your coworker behaviour to a system that does not work like a coworker, which is where the real risk starts.