Finding Your Tribe in the Age of AI

The AI Everyone Talks About vs. the AI Nobody Talks About

When people talk about AI and human connection, the conversation almost always goes in one direction: AI is replacing it. Chatbots instead of therapists. Companion apps instead of friends. Generated content instead of human expression. The narrative is that AI is the latest technology to isolate us further, to give us synthetic substitutes for the real relationships we increasingly cannot form.

This narrative is not wrong about the trend. AI companion apps like Replika and Character.ai have millions of users, many of whom describe their AI chatbot as their closest relationship. The loneliness epidemic that the Surgeon General identified is being answered, in part, by artificial relationships. That is genuinely concerning.

But this narrative is incomplete. It assumes that AI can only substitute for human connection. It ignores the possibility that AI can facilitate it — that the same technology capable of simulating a friend could be used to help you find a real one.

How AI Currently Mediates Human Interaction

AI already controls most of the human interaction that happens online. It is just that almost all of it is designed to keep you alone with your screen.

TikTok's recommendation algorithm is perhaps the most sophisticated AI system ever deployed for consumer attention capture. It learns your preferences with eerie precision, creating a personalized infinite scroll that is nearly impossible to put down. The AI does not connect you with people. It connects you with content designed to keep you watching. Other humans exist only as content creators — performers for your algorithmic feed. You are never prompted to talk to any of them.

Instagram's Explore page uses machine learning to surface content optimized for engagement. The AI analyzes your behavior — what you like, what you linger on, what you share — and builds a model of what will keep you scrolling. Again, the AI mediates your relationship with content, not with people. Other humans are sources of consumable media, not potential conversation partners.

Dating apps use AI and machine learning for matching, but the matching is based on surface-level attributes: photos, stated preferences, swipe behavior. The AI is optimizing for mutual physical attraction and stated lifestyle compatibility, not for intellectual or creative resonance. And the apps are designed to maximize engagement (keeping you swiping) rather than to maximize successful connections (getting you off the app).

LinkedIn's "People You May Know" uses AI to expand your professional network, but it is based entirely on the social graph: shared connections, shared employers, shared schools. The AI does not know what you actually think about. It knows who you have been proximate to.

In every case, AI is being used to keep you consuming, scrolling, and engaging with a platform — not to help you find the specific humans who share your intellectual and creative DNA.

What If AI Read What You Create Instead of What You Consume?

Here is the alternative that almost nobody is exploring — and one we explore further in going beyond the algorithm: instead of using AI to analyze what you consume (and then feed you more of it), use AI to analyze what you create (and then connect you with people who create similar things).

What you consume is a weak signal. You scroll past thousands of pieces of content per day. You "like" things impulsively. Your consumption patterns reflect habit, boredom, and algorithmic manipulation as much as they reflect genuine interest.

What you create is a strong signal. The essay you spent three hours writing. The notes you took while reading a book that fascinated you. The journal entry where you worked through a question that had been bothering you for weeks. The sketch, the poem, the half-finished project in your notes app. These are not impulse actions. They are expressions of what you genuinely care about, think about, and return to.

If an AI system can read what you create and understand the concepts, themes, and questions embedded in your output, it can build a map of your intellectual life — a semantic fingerprint that captures not just your topics of interest but the specific angles, combinations, and depths that make your thinking unique.

How Semantic Embeddings Work (Without the Jargon)

The technology that makes this possible is called semantic embeddings, and while the name sounds intimidating, the concept is surprisingly intuitive.

Imagine that every idea, sentence, or paragraph you write can be placed as a point in an enormous multidimensional space. In this space, ideas that are similar in meaning are located near each other, even if they use completely different words. "The architecture of living systems inspires better building design" and "biomimicry is transforming how we think about structural engineering" would be very close to each other in this space — not because they share keywords, but because they express similar concepts.

An embedding model is an AI system that converts text into these spatial coordinates (called vectors). A typical embedding model maps text into a space with hundreds or even thousands of dimensions — far more than we can visualize, but mathematically straightforward to work with. The model has been trained on enormous amounts of human text and has learned, through that training, to position semantically similar ideas near each other.

When your notes are processed through an embedding model, each chunk of text gets its own position in this conceptual space. The resulting collection of positions — your embedding profile — is essentially a map of your intellectual landscape. The clusters reveal your recurring themes. The outliers reveal your unique combinations. The overall shape reveals what kind of thinker you are.

Now here is the powerful part: you can compare one person's embedding profile with another's and measure how similar they are. Not at the keyword level ("you both mentioned 'design'"), but at the conceptual level ("you both keep circling back to the tension between emergence and control in complex systems"). This is a fundamentally deeper form of matching than any tag-based or keyword-based system can achieve.

From Embeddings to Tribe

Once you can measure conceptual similarity between people, you can do something that has never been possible before: automatically identify your tribe.

Your tribe is not the people who live near you. It is not the people who went to your school. It is not the people who work in your industry. It is the people who think the way you think — who are wrestling with the same questions, drawn to the same ideas, making the same unexpected connections between fields.

Before the internet, finding your tribe was mostly luck. You happened to sit next to the right person at a conference. You happened to find the right book in the right bookstore. You happened to take the right class with the right professor who introduced you to the right ideas and the right people. Brilliant minds spent their entire lives in intellectual isolation because the handful of people who shared their interests were scattered across the globe.

The internet was supposed to fix this, but it did not — because the internet organized people by social connections (who you know) and self-declared interests (who you say you are), neither of which is an accurate proxy for intellectual resonance.

AI-powered semantic matching can actually fix this. It can read what you have written, understand what you think about, find the other people on the planet who think about similar things, and put you in a room together. Not a room of 50,000 strangers, but a room of 15 people who share your specific combination of interests at a depth that keyword matching would never detect.

The Objections (and Why They Matter)

There are legitimate concerns about this approach that deserve honest engagement:

"Isn't this just a more sophisticated filter bubble?" Potentially, if done badly. The paradox of open communities shows that some filtering actually improves quality — but the key is matching on shared questions, not shared conclusions. If the matching algorithm only connects you with people who agree with you, it creates an echo chamber. The key design choice is to match on shared interests and questions, not on shared conclusions. Two people who are both fascinated by the ethics of genetic engineering may have completely opposite views on the subject — but they will have a far more interesting conversation than two people randomly assigned to debate it.

"Do I want AI reading my private notes?" This is a serious concern. The answer has to be that the analysis happens with full user control and transparency. You choose what to share. You can see exactly what concepts the AI extracted. You can delete your data at any time. And the raw content of your notes is never shared with other users — only the conceptual patterns extracted from them. Privacy-respecting design is not optional here; it is foundational.

"Won't this just be used by a certain type of person?" Probably, at first. People who write a lot, think a lot, and have extensive notes will get the most out of semantic matching. But the vision is broader: as voice-to-text, image analysis, and other AI capabilities improve, the input does not have to be text. It can be your sketches, your voice memos, your photographs — any creative output that reveals what you care about.

The Vision

The dominant AI narrative is about replacement: AI that writes for you, thinks for you, keeps you company so you do not need other humans. There is a different narrative that deserves more attention: AI that understands you well enough to help you find the humans who understand you too.

This is the vision behind kndred. Not AI as a substitute for human connection, but AI as a bridge to it. You bring your ideas — in the form of your writing, your notes, your creative output. The AI reads them, understands the conceptual landscape they describe, and connects you with the other people whose landscapes overlap with yours. Then the AI steps back, and the humans take over.

The technology is a means, not an end. The end is the conversation that happens when two people discover they have been independently thinking about the same things. The end is the message that says "I have been obsessed with this exact question for three years and I have never met anyone else who cares about it." The end is the experience of being understood.

AI does not have to make us lonelier. Used thoughtfully, it can do something the internet has always promised but never delivered: help every person find their tribe.