Your AI already knows you. Veram introduces you.
For years you’ve asked an AI the things you don’t put in a profile: whether to take the job, how to say the hard thing, what to cook for someone. Veram turns what it has noticed into an honest portrait of you, each claim saying how your AI knows it, and uses it to introduce you to people who fit.
- 01Your AI writes your portraitFrom what it already knows about you. Every claim says how it knows, and roughly when.
- 02You make it trueFix anything it got wrong, answer seven things no AI can know, add a photo.
- 03We introduce youA few people at a time, within the miles you choose. Where you rhyme, and one honest rub. No names or photos yet.
- 04You choose to meetSay meet to see who they are. If they say it too, your conversation opens.
THE DETAILS
1. Your AI writes your portrait
You pick ChatGPT, Claude or Grok; it opens with Veram’s prompt already filled in. With memory or chat-history reference switched on, it writes a portrait from what it remembers, the way people who know you would put it: how a friend would introduce you, what a date notices in the first hour, what your friends would say, what lights you up, what you’re like to live with, where you’re heading, and what takes getting used to. Every line has to pass one test: would you be happy for a date to read it aloud? It leaves out your work and how you use your AI. Each claim says how it knows it and roughly when, such as “several chats about trips, 2025”. If it can’t see your history, it says so instead of inventing anything.
2. Veram keeps the portrait, not the conversation
You paste the portrait into Veram. We read it once and discard the text. Before anything is stored, a privacy gate removes anything about other people in your life, and anything about money, health, religion, politics or sexual orientation. Veram never connects to your AI account and never sees your chats.
3. You check it, and add what no AI can know
Every claim is right by default; flag the ones that are partly or plainly wrong. Say who you’re looking for and where you are. Then seven quick answers that matter more than any trait: kids, relationship shape, smoking, how you handle a disagreement, what a good week together looks like, where you’ll be living, and where the next two years are taking you. Add at least one photo, up to six, hidden until someone chooses to meet you. If you like, verify with a live selfie.
4. A deep profile, sharpened by your own AI
From all of that, Veram places you on a few dozen scales: what you value, where your life is going, temperament, how you are with someone, your daily rhythms, and your style in a disagreement. You see every scale. One more prompt lets your own AI score them from its full memory, and you can answer the ones nobody could tell. Two private scales about steadiness and closeness are off unless you switch them on; they’re never quoted or shown to anyone.
5. Matching that explains itself
First, hard lines in both directions: age and gender preferences, the miles you choose, kids, smoking. Then a score weighted toward what relationship research actually supports: shared values and life direction; compatible ways of handling conflict; needs for closeness and space; rhythms that could share a week. A matchmaker model reads your shortlist, and each introduction comes with a written report: where you rhyme, with a line from each of you, and the one place you’d rub. You get at most three at a time: two best fits and one a little outside your pattern.
6. Choose to meet
Before you choose, you see a match percentage, the shared ground and the rub. Never a name or a photo. Saying meet is the reveal: their name, photos and full report. If they say meet too, a private thread opens, live, with three questions to start from.
Your portrait and every introduction are free to read. A reveal costs $9.99, or 3 for $24.99, or 5 for $39.99; the first 20,000 members get their first reveal free. Saying meet back to someone who already said meet to you is always free. Veram opens first in New York City. If a reveal doesn’t become a conversation within 7 days (both of you writing), one reveal is added back to your account automatically.
The science, honestly
No algorithm can promise chemistry. When researchers gave machine learning more than 100 measures of people’s traits and preferences before speed dates, it could predict who tends to be liked, but not who would like whom (Joel, Eastwick & Finkel, 2017). A major review of online dating found no compelling evidence that matching algorithms work better than other ways of pairing people (Finkel et al., 2012). What does predict how relationships go is mostly how two people experience each other once they’re together (feeling their partner is committed, appreciation, how much they fight), then each person’s wellbeing and security with closeness (Joel et al., 2020, across 43 studies), and how a couple handles conflict and comes back from it (Gottman’s work).
We wrote the longer version, with every source, in the Journal: what actually predicts a good relationship, and what doesn’t.
So Veram does three things it can actually stand behind: it avoids introductions that were bound to fail, it explains every introduction with evidence from both people, and it learns from which introductions led to a real meeting.
Try it on someone you know
Not ready for strangers? Compare portraits with a partner, a friend or a sibling and get one report on the two of you: where you rhyme, where you don’t, and three questions to ask each other tonight.