we're building the most sophisticated matching platform ever made.

there's a moment in every dating app where you hit the bio screen. the cursor blinks. you're supposed to distill yourself into 500 characters. maybe you write "love hiking and tacos" because that's what everyone writes. maybe you stare at it for ten minutes and close the app.

that moment is where most dating technology begins. and it's exactly where ours doesn't.

we are larry match. and we're building a matching platform that doesn't ask you to describe yourself at all.

the problem with self-reporting

here's something the dating industry doesn't talk about: the data that powers your matches on every major platform is self-reported. you fill out prompts. you pick photos. you write a bio. the algorithm uses that to decide who you should meet.

the issue isn't that people lie (though they do). the issue is that people genuinely don't know. ask someone what they value in a partner and they'll give you an answer. watch how they actually behave and you'll get a completely different one. psychologists have a name for this gap. it's called the intention-behavior gap, and it's one of the most replicated findings in social science.

every dating app on the market builds its matching engine on top of this gap. they're optimizing on what people say they want, not what they actually respond to.

we decided to try something different.

behavioral truth

what if, instead of asking someone to describe themselves, you could observe how they already behave? not in a lab. not through a questionnaire. just... in the wild. the way they talk to strangers. the things that make them laugh. the causes they care about. how they argue. what they share when nobody's grading them.

social media is the largest behavioral dataset in human history. billions of people expressing themselves daily with no romantic incentive. no one crafts a post thinking "this will help me find a partner." that's exactly what makes it valuable.

we built an AI system that reads this behavioral data and constructs a personality profile. not a simple one. not "introvert or extrovert." we're talking about a multi-dimensional model that captures how someone thinks, what they value, how they handle conflict, what kind of humor they use, whether they're a morning person, how they relate to money, what their attachment patterns look like.

dozens of dimensions. each one weighted differently depending on what kind of relationship you're looking for.

from numbers to nuance

every matching system in the industry works the same way at its core: turn people into numbers, compare the numbers, sort by score. and honestly? that part works. math is good at narrowing a pool of thousands down to a few dozen people who are plausible on paper.

the problem is what happens next. on every other platform, the math is the end of the line. the algorithm hands you a ranked list and says "good luck." maybe the top result scores a 94 and the fifth scores an 88. you'll never know why, and neither does the system. it did the arithmetic and clocked out.

we use that same mathematical narrowing. it's a genuinely good first step. but we treat it as exactly that: a first step. once the numbers have done their job and surfaced a strong candidate pool, something else takes over. an AI that's been paying attention.

not paying attention to your score. paying attention to you. what you responded to last time. what you told larry you were looking for (and what your behavior suggested you actually want). whether you've been skipping the same "type" for two weeks. whether your taste is shifting in ways you haven't articulated yet.

the AI reviews the candidates the way a great human matchmaker would. not by who ranks highest on a spreadsheet, but by who fits you right now, today, given everything it knows about how you've been showing up. the math finds the pool. the intelligence does the curation.

that's the leap most people don't realize is possible. going from "here are your top matches by score" to "here's someone i think you need to meet, and here's why." the first one is a search engine. the second one is a matchmaker.

a matchmaker, not a filter

the other thing we did differently is build a character.

his name is larry. he's your matchmaker. you don't swipe. you don't browse. you talk to larry, and he talks to you. when he thinks you should meet someone, he tells you why in plain language. not a compatibility percentage. something closer to: "she argues the same way you do, and you both think you're funnier than you are."

larry gets to know you over time. and he remembers everything. the more you talk to him, the better he gets at understanding what you're actually looking for, even when you're not sure yourself.

he's not the only thing we built. there are other layers to the experience that we're not ready to talk about yet. what we will say is that larry is just the beginning of what users encounter, and the product has a way of surprising people the longer they use it.

what we're not

we're not a swipe app with an AI sticker on it. there are a lot of those now. companies that bolt a language model onto an existing dating interface and call it "AI-powered matching." that's like putting a racing stripe on a minivan.

real AI matching means the intelligence isn't a feature. it's the foundation. every part of the product, from how profiles are built to how matches are scored to how conversations happen, runs through the AI layer. there's no version of larry match without the AI. remove it and there's nothing left.

we're also not trying to replace human judgment. larry doesn't decide who you date. he introduces. he explains. he sometimes nudges. but you make every call. the AI's job is to surface people you'd never find on your own, people who don't look like your "type" on paper but share the behavioral patterns that actually predict compatibility.

the weight of small signals

one of the things that surprised us most during development is how much signal lives in the small stuff. not the big declarations (political affiliation, religion, whether you want kids). those matter, obviously. but the texture of compatibility, the stuff that determines whether a first date leads to a second, lives in subtler places.

how someone uses humor. whether they're a generous communicator or a transactional one. how they respond when they're frustrated. whether they engage with the world out of curiosity or obligation. these aren't things you'd put in a dating profile. most people couldn't articulate them about themselves. but they show up in behavior. consistently.

our system tracks dozens of these signals. each one is weighted based on what kind of relationship you're looking for: a casual connection weights different signals than a long-term partnership. the weights aren't static. they adjust as the system learns from outcomes. what actually works, not what we assumed would work.

what's possible

our founding team spent years engineering processes before starting this company. optimizing routes and workflows. the core principle was always the same: the system knows things the individual participants don't. a single subway rider can't see the whole network. but the network can see every rider (unless you're the NYC MTA lolz).

matching works the same way. you can't see the full landscape of people who might be right for you. but a system that understands personality at depth, across a large population, with the ability to learn from every interaction? that system can find connections that no profile-browsing, no swiping, and no algorithm-tweaked questionnaire ever could.

we think we're building the most sophisticated matching platform ever made. that's a big claim. but when you look at what the industry has been doing for the last decade, the bar is lower than you'd expect.

most of what exists is filtering. you set preferences, the app filters people who match those preferences, and you swipe through the results. that's not matching. that's search.

matching is prediction. it's answering the question: if these two people sat across from each other, would something happen? and the only way to answer that question well is to understand both people at a level that goes far beyond what they'd tell you about themselves.

that's what we're trying to build. we're not there yet. but the early results suggest we're pointed in the right direction.

check back in may – we move fast.

— calum dew CEO, larry match