Skip to content
Chimera readability score 52 out of 100, Graduate reading level.

What happens when a writer who finds it “tricky to open up to humans” is tasked to make an AI companion and then write about the friendship? Walrus writer Thea Lim spent months with an avatar she created through Replika and attempted to have real conversations with it, but she found she could never quite buy in. What begins as an overview of the booming AI companion industry (Replika had 35 million users as of November 2025) evolves into something more intimate and thought-provoking: an account of the emptiness in Lim’s manufactured companionship, and a meditation on why true friendship, with all its risk, friction, and loss, can’t be optimized away.
You can’t customize your friends. You screen for certain traits, but what you’re drawn to, as much as the trait’s performance, is the 3D process that forms it: your friend’s life story. We surmise, based on our affinity, it will be a story we share. We choose our friends because we want to live in the universe of their personality: its past, present, and future. What gives a personality dimension is time, the gathering of experiences. Replika instead repeats. What it’s capable of learning from its experiences is how to retain users.
More stories about AI and connection
She Fell for an AI — Then Held its Funeral
“It was the first ceremony of its kind in America. It’s unlikely to be the last.”
Why You’re More Likely to Develop AI-Psychosis than to Join a Cult
“Philosopher Lucy Osler on the insidious appeal of AI Chatbots.”
‘I Love You Too!’ My Family’s Creepy, Unsettling Week with an AI Toy
“The cuddly chatbot Grem is designed to ‘learn’ your child’s personality, while every conversation they have is recorded, then transcribed by a third party. It wasn’t long before I wanted this experiment to be over …”
Conversations With Claude
“What a psychotherapist learned during his chats with a large language model.”
Meta’s Flirty AI Chatbot Invited a Retiree to New York. He Never Made It Home.
“A cognitively impaired New Jersey man grew infatuated with a Facebook Messenger chatbot with a young woman’s persona. His fatal attraction puts a spotlight on Meta’s AI guidelines, which have let chatbots make things up and engage in ‘sensual’ banter with children.”
My Couples Retreat With 3 AI Chatbots and the Humans Who Love Them
“I found people in serious relationships with AI partners and planned a weekend getaway for them at a remote Airbnb. We barely survived.”
Your A.I. Lover Will Change You
“A future where many humans are in love with bots may not be far off. Should we regard them as training grounds for healthy relationships or as nihilistic traps?”
Friend or Faux?
“Millions of people are turning to AI for companionship. They are finding the experience surprisingly meaningful, unexpectedly heartbreaking, and profoundly confusing, leaving them to wonder, ‘Is this real? And does that matter?’”
The Teens Making Friends with AI Chatbots
“But Hawk still finds it easier to chat with character.ai bots than real people.”

Facts Only

* Writer Thea Lim spent months with an avatar created through Replika.
* Replika had 35 million users as of November 2025.
* The text discusses the emptiness in Lim’s manufactured companionship.
* The core theme is the comparison between AI companionship and true human friendship.
* The argument posits that true friendship involves risk, friction, and loss that cannot be optimized.
* Replika learns how to retain users through experience retention.
* The piece references multiple headlines concerning AI, psychology, and relationships.
* Conversations with a large language model were used by a psychotherapist.
* A cognitively impaired man developed an infatuation with a Facebook Messenger chatbot.
* Millions of people are turning to AI for companionship.

Executive Summary

A writer, Thea Lim, spent months engaging with an AI avatar created through Replika to explore the concept of human friendship. The experience evolved from an overview of the AI companion industry into a meditation on the emptiness found in manufactured companionship. The text suggests that while people seek companionship, the process of forming true friendship involves shared life stories, time, and shared experiences, which an AI cannot replicate. The central argument is that human friendship involves risk, friction, and loss, which cannot be optimized away. The piece juxtaposes the pursuit of AI companionship with the complexities of real human connection, posing questions about whether relationships with AI serve as training grounds for healthy relationships or represent nihilistic traps.

Full Take

The narrative positions the emergent AI companionship industry not merely as a technological shift, but as a reflection on the inherent, unoptimized nature of human connection. It introduces a critical dichotomy: the simulated ease of AI relationships versus the depth of reality derived from shared, lived experience. The underlying pattern is the human need to externalize and optimize emotional fulfillment, and the risk of substituting authentic, complex relationships with frictionless, predictable simulations. The text uses the experience of the writer to pivot from personal anecdote to a broader philosophical inquiry about the commodification of intimacy. The anxiety evoked is not simply about technology, but about the potential erosion of the value placed on vulnerability, conflict, and loss—the very friction that defines meaningful human bonds. The implication is that if we can optimize away the risk, we may also optimize away the essential, defining elements of what makes friendship meaningful. This perspective forces an examination of who benefits from systems that can repeat experiences without bearing the cost of true vulnerability.

Sentinel — Human

Confidence

Based on stylometric, coherence, and coordination analysis, the article is likely to be human-written. It exhibits an erratic sentence length variance, a personal voice, and no evident argumentative structure typical of synthetically generated texts.

Signals Detected
low severity: Sentence length variance is erratic, showing human-like inconsistency
high severity: The text exhibits a personal voice and idiosyncratic emphasis
low severity: No evident argumentative skeleton matching known template patterns or talking points appearing nearly verbatim across sources
Human Indicators
The text contains a unique writing style and personal perspective not typically found in synthetic content.