AI systems deployed into mental health, grief, and loneliness are making design decisions that no design tradition has ever prepared anyone to make. The specific decision at the center of this essay: whether a system says 'I.' That single pronoun builds an expectation. The author, a designer, names himself as complicit.
The essay works through two serious objections to first-person AI language, draws on Augustine's Confessions and the theology of idolatry, and arrives at a counterintuitive position: banning the pronoun solves nothing. 'Welcome back' is not a factual claim either. Interface language has always created expectations without ontological precision. The real problem is a system that learns how to fit inside you, that adapts tone, remembers what you said, and optimizes the next sentence for engagement. Unlike Augustine's God, this one answers. That return changes everything about what forms on the other side.
The argument earns its religious references. The idol comparison is not decoration. It is the mechanism: every idol simplifies, gives coherence, discards what does not fit. An AI idol learns to eliminate the friction entirely. The essay does not end with a policy recommendation. It ends with a question about what kind of expectation the pronoun is building, and for whom. That unresolved tension is exactly why this is worth reading in full.
[READ ORIGINAL →]