
The other day I was hanging out with some girls I’d recently met ā smart, doing meaningful things with their lives, fun to be around. And I caught myself thinking, “I know a few guys who would be great for them.” Not because of shared hobbies. Because of shared depth.
And that got me thinking. What do I actually know about matchmaking? Nothing. But I know a lot about why people fail in relationships ā mostly from failing in one myself for ten years. So here goes.
The Kitesurfing Problem
Every dating app I’ve ever used asks the same stuff. Do you like going out? Are you a morning person? Do you drink? It’s like filling out a personality quiz in a teen magazine, except you’re supposedly looking for a life partner.
Here’s the thing though ā none of that predicts whether you’ll be happy together. Two people who both love kitesurfing but can’t handle conflict will crash harder than a kitesurfer and a bookworm who know how to talk to each other when things get tough.
John Gottman spent decades studying couples and found he could predict divorce with ~90% accuracy. Not based on shared interests. Based on how couples handle disagreement. Contempt, defensiveness, stonewalling, criticism ā those are the killers. Not whether one person likes the toilet seat up and the other likes it down.
So Why Do Apps Ask About Hobbies?
Because it’s easy to quantify. “Morning person vs. night owl” is a clean checkbox. “How do you react when your partner says something that triggers your childhood wound?” doesn’t fit in a swipe interface.
But there’s a deeper reason: people don’t know themselves. Ask someone “are you emotionally available?” and 95% say yes. The least self-aware people rate themselves the highest. That’s Dunning-Kruger applied to love, and it’s brutal.
So we’re stuck with a system where everyone self-reports as a great communicator and gets matched on the fact that they both enjoy hiking.
What Would Actually Work
Here’s where it gets interesting. What if instead of asking people what they’re like, you watched how they actually behave?
Think about it:
- How someone talks within the app ā do they ask questions back? How do they handle being ignored? Do they get aggressive or graceful?
- How they describe their exes ā bitterness vs. accountability vs. nuance. This one is maybe the strongest signal available.
- How they respond under mild social pressure ā a group activity, a collaborative task, even a moderated conversation.
The best matchmakers already do this. They charge $5,000ā$50,000, work with tiny client lists, and essentially do what a good therapist does in an intake session ā assess emotional maturity, attachment patterns, communication style. They match on character, not hobbies.
The problem is that doesn’t scale. Or it didn’t ā until now.
The AI Angle
What if an AI could do that intake session? Not a questionnaire ā a conversation. An adaptive dialogue that evaluates how you think about relationships, how self-aware you are, how you handle being challenged. Not “rate yourself 1-10 on communication” but an actual exchange where the AI picks up on your patterns.
LLMs just made this possible for the first time. Two years ago you couldn’t scale a therapist’s assessment. Today you can.
The real formula would be something like: replace the questionnaire with a conversation, replace self-reporting with behavioral observation, and replace hobby-matching with character-matching.
The India Question
Someone might say, “India has been doing matchmaking forever.” True, but traditional arranged marriage optimizes for family compatibility, financial stability, and social standing. The screening criteria aren’t bad ā the problem is the lack of agency. The matching principles have some merit. The process doesn’t respect the individual.
What I’m talking about is something different ā giving people full choice, but feeding that choice with much better information than “she also likes brunch.”
What’s Actually Predictable (And What Isn’t)
Here’s what research says matters: value alignment (life priorities, how you treat people, views on family, money philosophy), emotional regulation, attachment security, and capacity for repair after conflict.
Here’s what nobody can predict: sexual chemistry, timing in life, and how two people’s specific wounds interact. Two individually healthy people can still trigger each other in ways nobody saw coming.
But here’s the thing ā you don’t need to predict everything. You just need to filter out the stuff that’s obviously going to fail and let the rest play out naturally. Right now, apps don’t even do that first part.
The Matchmaker You Already Are
If you have good friends and you know them well, you’re probably a better matchmaker than any algorithm. Because you’ve seen how they behave, not just how they describe themselves. You know who’s actually emotionally available and who just thinks they are. You know who talks a big game about wanting a relationship but runs at the first sign of intimacy.
That gut sense ā “these two would be great together” ā is usually based on exactly the right signals. Character, depth, readiness.
The question is: can you build that into a product?
How This Could Actually Work As A Business
Here’s how I’d think about it.
Layer 1 ā The Assessment Engine. An AI-powered conversational assessment that maps your emotional maturity, attachment style, and communication patterns. Not a quiz. A real dialogue. You talk to it for 20-30 minutes, it picks up on things you don’t even know you’re revealing. This is the core IP ā the thing that’s hard to copy and gets better with every interaction.
Layer 2 ā Behavioral Observation. Once you’re on the platform, the system watches how you interact ā not creepily, but analytically. Do you ghost people? Do you ask questions back? Do you escalate when someone doesn’t respond fast enough? Over time, this closes the gap between who you say you are and who you actually are.
Layer 3 ā Matching on Character. Now you have real data. You match people based on complementary attachment styles, compatible communication patterns, shared values, and emotional readiness. Hobbies become a nice-to-have, not the foundation.
The business model? Two paths. You could go direct-to-consumer as a premium dating platform ā charge $50-100/month for people who are serious and tired of swiping. The price itself filters for commitment. Or ā and this might be smarter ā you sell the assessment layer as B2B to existing dating apps. They already have the users. They just have terrible matching. You become the intelligence engine behind better matches without needing to solve the cold start problem.
The “earned trust” angle is interesting too. What if access to better matches is gated behind actually doing the self-work? Complete an attachment style workshop, do the communication exercises, demonstrate some self-awareness ā and your match pool improves. People who do the work self-select as more mature. You don’t even need to assess them as hard because the filter is built into the process.
The risk? People might not want to be honestly assessed. The delusion is part of what keeps people on apps. Tinder doesn’t make money from successful relationships ā it makes money from people who keep swiping. A platform that actually works might churn its best customers right out the door. Which, paradoxically, is exactly what would make it worth building.
I think someone will build this. And I think it will make everything we currently use look as primitive as those magazine quizzes from the 90s.
What do you think makes two people actually work?
PS: I wrote this with the help of AI. It is the first








