A matchmaker, not a slot machine.
Consumer software · Dating
Pear interviews you like a matchmaker would, then hands you a short deck of people it can explain — why this person, what you'd have in common, what to open with. Swiping is still there. It's just no longer the whole product.
- Stage
- Pre-launch — v0.1.0, internal builds
- Platforms
- iOS and Android, one React Native codebase
- Stack
- Expo · React Native · Supabase · Gemini
- MTHR role
- Founded, built and operated in-house
Why it exists
Dating apps optimised themselves into a volume game: infinite decks, no explanation, and a business model that quietly prefers you stay single. The people who do best on them are the people who need them least.
Pear inverts the ratio. Fewer candidates, each with a reason attached, plus the part every app leaves out — help with the conversation after the match, which is where most of them actually die.
Onboarding is the interview
The first screen asks how you'd rather do this: talk to Pear, or fill in the slides. Take the conversation and it interviews you — name, birthday, location, photos, who you're looking for, your dating style, love languages, social energy, a personality read and a set of scales — filling the profile through the conversation and handing you a review screen at the end. Take the slides and you get the same profile without the small talk.
What the AI actually does
- Conversational onboardingPear fills in your profile through function calls as you talk, so the interview isn't a transcript sitting next to an empty form — it is the form.
- Deck curation with reasonsCandidates are scored and ranked, and each one arrives with a written summary: why you two, what you share, and openers that fit.
- Message coachingDraft suggestions inside every chat, plus a "how would they hear this" read before you send something you'll regret.
- Safety scanOpt-in scanning of received messages for concerning patterns, toggled from your profile — off by default, yours to turn on.
- Date planningStructured plans for a specific match: venue, cost, vibe — the logistics that stall the jump from chat to meeting.
- Memory and learned traitsPear remembers what you've said you want and infers taste from how you actually swipe — and both are visible and editable on your profile, rather than a black box guessing at you.
The rest of the app
Around the deck: an explore surface, a chats list with per-match threads, a dedicated assistant tab, a recommendations view, match pages, hobby pages, a date planner and a scheduler, profile and edit-profile, reviews after a date, and report and block flows one tap from any profile. Terms, privacy and community guidelines ship inside the app rather than behind a link to a website.
Every match arrives with an argument for itself.
Under the hood
| App | Expo SDK 54, React Native 0.81, React 19, TypeScript in strict mode, file-based typed routing |
|---|---|
| Offline-first | Deck, matches, threads, profile, memory and learned traits each have their own local store on AsyncStorage; the whole flow runs against mock candidates with no backend and no API key |
| Backend | Supabase — Postgres, Auth and photo storage, with row-level security on profiles, likes and passes, matches, messages, reports, blocks, events, reviews, scheduled dates and favourites |
| Cloud sync | Every cloud call short-circuits cleanly when Supabase isn't configured, so the offline path is never a second-class code path |
| AI | Gemini 2.5 Flash, called through a Supabase Edge Function so the key never ships inside the app binary |
| Trust and safety | 18+ age gate in onboarding, in-app account deletion, and a report/block pipeline writing to RLS-protected tables |
| Admin | A separate web app for roadmap and moderation |
Where it is now
- DoneThe whole product loop Onboarding, curation, swiping, matching, chat, coaching and date planning all work end to end.
- DoneLaunch-readiness engineering AI key moved server-side, RLS schemas for every table, account deletion, age gate, report and block flows.
- DoneFirst Android build v0.1.0 packaged for internal installs.
- NextProduction Supabase and TestFlight Point at the production project, apply the migrations, deploy the AI proxy, then the first iOS build goes out to testers.