January 15, 2026

What your masi knew that algorithms don't

Every Indian family has one. The aunt — usually on your mother's side — who knows everyone's business and considers it her sacred duty to deploy that information at weddings. She remembers who married whom in 1987. She knows which families had "issues." She has opinions about the Mehta boy that she will share whether you ask or not.

We named our company after her. Not as a joke. Because she was right more often than any of us want to admit.

The intelligence network you did not appreciate

Your masi operated the most sophisticated recommendation engine in your community, and she ran it entirely on social data that no app has ever captured.

She knew that the Sharma family was warm and welcoming because she had been to their house for chai and seen how they treated guests. She knew that Raj was a good boy not because of his LinkedIn but because she had watched him touch his grandmother's feet at every family function without being prompted. She knew that Priya was sharp and independent because she had seen her navigate her father's illness with grace while finishing her CA exams.

This is not gossip. This is contextual intelligence. And it is precisely what dating apps strip away when they reduce a person to six photos and a 300-character bio.

What checkboxes miss

Open any matrimonial site. The filters are the same ones they have had for twenty years. Height. Income. Caste. Manglik status. Education. City. These are not useless data points, but they are the skeleton without the flesh.

A checkbox cannot tell you that someone lights a diya every Thursday evening but has not been to a mandir in three years — and that this combination makes perfect sense to them. A dropdown menu cannot capture that someone is vegetarian at home out of respect for their mother but eats everything when traveling. A text field asking "describe yourself" will never surface the fact that someone spends every other weekend driving to Edison to have dal-dhokli with their aging foi.

These are the details that determine whether two people can actually share a life. And they only emerge in conversation.

Your masi never asked someone to fill out a form. She sat with them. She asked about their family. She watched how they spoke about the people they loved. That is the intake process we modeled Masii after.

Context is not a feature — it is the product

When we built Masii, we spent a long time studying what traditional matchmakers actually did well. The answer was not astrology or caste matching or even the extensive networks, though those helped. The answer was context.

A good matchmaker in Rajkot did not just know that two families were both Jain. She knew that one family was Deravasi and the other was Shwetambar Murtipujak. She knew that the boy's family did upvas on paryushan but was relaxed the rest of the year, while the girl's family was strict about Jain dietary practices year-round — no root vegetables, no eating after sunset. She knew that this difference, invisible on any matrimonial profile, would become a daily negotiation in a shared kitchen.

This is what context means. Not broad categories but specific, lived details that predict how two people will navigate the ordinary Tuesday evenings that make up a marriage.

AI can learn this. Not through checkboxes, but through the same method your masi used — conversation. When someone tells Masii "I am cultural but not religious," we do not just tag them as moderate and move on. We ask what that looks like. Does cultural mean you celebrate Navratri with full garba for nine nights? Or does it mean you make kheer on Diwali and that is about it? Both are fine. They are just different, and the difference matters.

The things people reveal when they are not performing

Here is something your masi understood intuitively that the tech industry has been slow to learn: people reveal more about themselves in flowing conversation than in structured forms. A form invites performance. A conversation invites honesty.

When someone fills out a matrimonial profile, they write what they think a good match wants to hear. "Family-oriented. Enjoys traveling. Looking for someone with traditional values and a modern outlook." This describes approximately forty million people and distinguishes no one.

When someone talks to Masii about their last relationship, about what went wrong, about what they are not willing to compromise on — the real picture emerges. They mention that they cannot be with someone who does not get along with their sister. They mention that they need someone who understands why they send money home every month without being asked. They mention that they want a partner who will argue with them in Gujarati when they are upset because somehow the emotions only come out right in the mother tongue.

These are not things people put on profiles. They are things people say when they feel heard.

Old wisdom, new tools

We are not replacing your masi. Honestly, no one could. She has forty years of accumulated social intelligence and a memory for family trees that would put a genealogy database to shame.

But most of us do not live in communities where that network functions anymore. You are in San Francisco or Toronto or London. Your parents' network extends to other families who moved abroad in the same decade, and maybe those families have kids your age, maybe they do not, and maybe those kids are compatible, but probably the sample size is too small for a good match.

What Masii does is apply the same methodology your masi used — deep contextual understanding through conversation — at a scale she could never reach. She knew two hundred families. We can know twenty thousand. She could hold the nuances of her community in her head. We can hold the nuances of a diaspora.

The warmth, the intuition, the ability to read between the lines when someone says "I want someone independent" but means "I want someone independent who will still come to Sunday lunch at my parents' house" — that is what we built. Not a database with filters. A listener with judgment.