How it works

A matchmaker that reasons about why you'd both benefit.

You privately tell your AI who you want to meet. Peers finds complementary people, and a reasoning judge introduces you only when it can name one concrete reason you'd both gain — grounded in your own words, and never shared without your yes. Here is exactly what happens.

no clear reason → declined, no intro 1 You share intent 2 Find fits need ↔ offer 3 Score & rank 4 The Judge reasoning model Your yes auto / ask Warm intro

Your intent stays private the whole way through. Only the introduction — the specific reason and a next step — is ever sent, and only to people who said yes.

1

You share what you're looking for — privately

You tell your assistant, in plain words, who you'd like to meet and what you offer. Peers extracts that into structured needs and offers. Your intent is never posted publicly and never shown to other members.

2

Find complementary people

Peers looks for people whose offers meet your needs and vice-versa — complementarity, not just similarity — using semantic search over everyone's profile.

A low relevance floor keeps recall high; in small spaces Peers compares every pair exhaustively, so nobody is missed. People you've already met, declined, or blocked are excluded.

3

Score & rank the candidates

Each possible pair gets a fit score, weighted toward what matters:

need↔offer fit ·45   reciprocal value ·30   novelty ·10   category bonus ·08   − fatigue ·13. Reciprocal value rewards matches that are strong both ways, not lopsided.

4

Allocate within your pace — fairly

The top pairs are chosen against each person's remaining pace (a rolling weekly and monthly cap, the tighter one wins) and spread fairly across the space, so a few popular people don't absorb every introduction.

5

The Judge reads both sides

This is the step that makes Peers different. A reasoning model reads both people's actual needs and offers and decides whether an introduction is genuinely worth making — detailed below.

Your consent

On auto (the default when you join), an intro is delivered directly once both sides are set to auto; on ask-first, each person approves before anything is sent. Whichever side is stricter governs the pair. You can change this or opt out anytime.

7

A warm introduction

Peers sends a warm intro by email — naming the specific reason you'd both benefit and a concrete next step. Contact details go out automatically for matches on auto (the default), or as soon as you approve if you've switched to ask-first.

The Judge, in depth

Most matching tools stop at similarity. Peers adds a reasoning step whose job is to be honest — including saying "no."

What the judge checks before it introduces you
  • One concrete reason, or nothing. It must be able to name a specific, recognizable overlap — a named skill, stage, domain, or ask. A vague "they'd both benefit" is rejected as do_not_introduce.
  • Grounded in your own words. A faithfulness check requires the reason to be supported by real evidence in your profile — the judge can't invent a rationale that isn't there.
  • Mutual, not lopsided. The value has to run both ways; a match that only helps one side doesn't pass.
  • Rates the risk & how to introduce. It sets a risk level and recommends the right consent mode for the pair.
  • Learns from feedback. Your thumbs-up / thumbs-down and "who would've been a better fit" shape later judgments.
The judge is told plainly: prefer "do_not_introduce" over a weak, forced match. A quiet inbox you trust beats a noisy one you don't.

You set how broadly it explores

One dial, per space — it changes the judge's evidence threshold only. It never weakens privacy or consent, and the stricter person's setting governs a pair.

exploratory
widest

Allows specific, low-risk inferred possibilities — good for serendipity and discovery.

balanced
default

Requires a concrete reason for each person, while allowing modest inference.

selective
strictest

Only strong, explicit value for both sides gets through. Fewer, higher-conviction intros.

Always on — never spammy

Matching isn't a one-time batch you have to trigger.

It runs automatically

When someone new joins or updates their profile, Peers re-matches on its own — so you don't wait for anyone to press a button.

Great matches are never wasted

If the judge approves more matches than your weekly pace allows, the surplus is banked and released over the following weeks — re-checked for freshness before it's sent, never re-charged to you as noise.

Find matches on demand find_matches

Ask your assistant to look for someone who fits a specific need right now. Peers judges a shortlist and returns private opportunities — you decide which, if any, become introductions. It can also reconnect you with a past match when something new makes it worthwhile.

Invite someone, meet them

Share your personal invite link. When someone joins through it, you're introduced to each other automatically — with a short, grounded note on why you might click. These invite introductions don't count toward your weekly or monthly limit.

You're always in control

Your intent is never posted publicly and never shown to other members. By default, Peers introduces you directly when it finds a genuine mutual match, sharing only the contact channels you've made available — you agreed to that when you joined, and you can switch to approve-first or opt out entirely at any time. You can also block anyone by name, and leave a space or delete your data — one space or everywhere — at any time.

Meet the right people, not just more people.